# Insights into interconnections between the shallow and deep systems from a natural CO2 reservoir near Springerville, Arizona

**Authors:** Elizabeth Keating, Dennis Newell, David Dempsey, Rajesh Pawar  
**Published in:** International Journal of Greenhouse Gas Control (2014)  
**DOI (version of record):** https://doi.org/10.1016/j.ijggc.2014.03.009  
**This copy:** Published version. Text extracted from the PDF at https://ddempsey.github.io/papers/2014-keating-insights-interconnections-shallow-deep-systems/keating2014-insights-interconnections-shallow.pdf; equations, tables and figure placement may be imperfect, so cite the DOI.  
**Notice:** Published version, reproduced under the CC BY-NC-ND 3.0 licence. © The Author(s). Version of record: https://doi.org/10.1016/j.ijggc.2014.03.009  
**Licence:** CC BY-NC-ND 3.0 http://creativecommons.org/licenses/by-nc-nd/3.0/  

---
International Journal of Greenhouse Gas Control 25 (2014) 162–172

Contents lists available at ScienceDirect

International Journal of Greenhouse Gas Control

j ourna l ho me page: www.elsevier.com/locate/ijggc

Insights into interconnections between the shallow and deep systems from

a natural CO2 reservoir near Springerville, Arizona

Elizabeth Keating a,∗, Dennis Newell b, David Dempsey a, Rajesh Pawar a

a Los Alamos National Laboratory, United States b Utah State University, United States

a r t

i c l e i n f o

Article history: Received

9 September 2013 Received

in revised form 19 February 2014 Accepted

20 March 2014

Keywords: Groundwater

protection Faults Natural

analog

a b

s t r a c t

If carbon dioxide (CO2) sequestration into deep geologic reservoirs is to be accepted by the public and environmental regulators, the possibility of upward leakage into shallow groundwater should be acknowledged

and those processes well-understood. Studies of natural CO2 reservoirs and their connec- tion

(or lack thereof) with the shallow subsurface is one way to explore these issues. A natural reservoir near

Springerville, Arizona has leaked CO2 to the surface along a fault zone for thousands of years, creating large

travertine deposits. In recent times, the CO2 leak rates have declined signiﬁcantly yet the shallow aquifer

is still highly enriched in CO2. In this study, using water level data and simulations we demon- strate

that the fault zone likely provides hydrologic communication between the shallow aquifer and the

deeper reservoir. It is reasonable to assume, therefore, that the source of the CO2 in wells completed within the fault zone is the deeper CO2 reservoir. We present water chemistry data to demonstrate the geochemical

impact of this CO2 on shallow groundwater quality. Interestingly, arsenic concentrations are

elevated, but other trace metals concentrations are not. Arsenic and chloride concentrations co-vary, suggesting

perhaps an external source of both elements rather than an in situ release of As due to CO2 attack on shallow aquifer sediments. Observations at this site demonstrate that hydraulic communica- tions

between shallow and deep layers and upward CO2 migration does not preclude long-term viability of

a substantial CO2 reservoir at depth. We present multi-phase ﬂow simulations to illustrate possible mechanisms

trapping the CO2 at depth. Collectively, these analyses show that some degree of upward CO2 leakage may not be necessarily incompatible with the overarching goals of sequestering CO2 and protecting

shallow groundwater. Published

by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/3.0/).

1. Introduction

Injection of anthropogenic carbon dioxide (CO2) into the sub- surface has been suggested as an important component of a multi-faceted effort to reduce greenhouse gases (DOE). Carbon dioxide has been successfully injected into depleted oil ﬁelds for decades, but demonstrations of successful CO2 injection into other deep reservoirs, such as saline aquifers, are fairly few and all rela- tively

recent (Korbol and Kaddour, 1995; Preisig and Prevost, 2011; Hovorka et al., 2006, 2011). Of course it is not sufﬁcient merely to inject the CO2; a successful sequestration project will have to demonstrate that almost all of the CO2 stays in the deep reservoir and, if any leaks upward, that it does not negatively impact shal- low drinking water resources. This paper uses observations at a

∗Corresponding author. Tel.: +1 5056656714. E-mail

address: ekeating@lanl.gov (E. Keating).

natural CO2 reservoir near Springerville, Arizona to shed light on both long-term storage and upward leakage. Natural-occurring gas reservoirs are distributed throughout the world and have been exploited primarily for methane, noble gases, and carbon dioxide. The fact that these reservoirs have existed for millions of years has been used to argue that injected CO2 can be sta- ble in deep reservoirs and that caprocks can maintain their integrity for long periods of time. Some natural reservoirs, however, do leak gases to the surface. Understanding the mechanisms of leakage and the impact of the leakage on shallow groundwater resources using a ‘leaky’ natural reservoir near Springerville, Arizona is the focus of this study. Additionally, we address the question ‘how is it possible

that a substantial volume of gas-phase CO2 can be stored underground in the presence of a large leakage pathway?’ Previous studies of the impact of CO2 sequestration on shallow groundwater resources have focused on two issues: (1) CO2 intru- sion into the shallow aquifer, which may lower groundwater pH and mobilize trace metals (Little and Jackson, 2010; Kharaka et al.,

http://dx.doi.org/10.1016/j.ijggc.2014.03.009 1750-5836/Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/3.0/).

E. Keating et al. / International Journal of Greenhouse Gas Control 25 (2014) 162–172 163

2010), and/or (2) brine intrusion, which would introduce salts and trace metals into the groundwater (Keating et al., 2010). The leak- age pathway could be an abandoned well, a fault, or a localized caprock breach. Leakage pathways could be natural or ‘activated’ by overpressurization due to injection. Simulation and natural analog studies have addressed the physical mechanism of brine intrusion (Birkholzer et al., 2009; Oldenburg et al., 2011; Keating et al., 2012). Oldenburg et al. (2011) concluded that if overpressurization is mod- erate and salinity is high, upward brine migration will be limited and a new steady state can be achieved with the fresh water/saline water boundary well below the bottom of the aquifer. However, if overpressurization is high or salinity is low, brine will migrate into the shallow aquifer. Keating et al. (2012) used simulations and contrasting observations at two natural analog sites to conclude that the geometry of the leaky pathway will affect the likelihood of brine intrusion. They concluded that narrow features (wellbores, narrow fault zones) are more likely to convey brine into the shallow aquifer than broad leaky features. Birkholzer et al. (2009) analyzed an idealized multi-layer aquifer/aquitard system without a vertical conduit and concluded that brine intrusion would be unlikely in such a setting. The

impact of brine and CO2 intrusion on water quality in the shallow aquifer has been the focus of many recent studies. Lab- oratory experiments where sediments are exposed to CO2 have established the potential for trace metal release (Little and Jackson, 2010; Smyth et al., 2009). Likewise, controlled ﬁeld experiments have established that CO2 release can cause increases in trace metal concentrations (Kharaka et al., 2010; Trautz et al., 2013, Peter et al., 2012). In most cases, the increase in trace metal concentrations was very subtle and concentrations did not approach or exceed drink- ing water standards. Some laboratory experiments (Carey et al., 2009) and natural analog studies (Keating et al., 2013) suggest that trace metal elevation will be temporary. As buffering reactions proceed, trace metals can precipitate or re-adsorb (Aiuppa et al., 2005). Apart from trace metals, saline brines can be a signiﬁcant threat to ground-water and NaCl will not be subject to signiﬁcant buffering. It is likely that each site will have a unique geochemical response to CO2 leakage.

2. Site description

The Springerville–St. Johns dome is a shallow (200–700 m) nat- ural CO2 reservoir near the southern edge of the Colorado Plateau (Moore et al., 2005), reported to contain as much as 445 billion m3

of CO2 (Stevens et al., 2005). The CO2 is closely associated with a relatively young volcanic ﬁeld (<3 Ma) (Moore et al., 2005). The reservoir is hosted in an alternating sequence of discontinuous sandstone, limestone, shale, and anhydrite layers of the Permian Supai Group. Below the Supai is fractured Precambrian granite basement; above are the Glorieta Sandstone and San Andres Lime- stone, which are regionally extensive aquifers. These, in turn, are capped by conﬁning layers (shales of the Moenkopi and Chinle For- mations). Northwest trending faults and related folds also cross the dome, with the Coyote Wash fault bounding the western margin of the dome. The Tuscon Electric Power (TEP) company wells produce groundwater from the Glorieta Sandstone (equivalent to the Coconino Sandstone). Regionally, this sandstone unit is the most productive unit of the C aquifer, the predominant groundwa- ter source in the Colorado Plateau/Little Colorado River basin of Arizona and western New Mexico (Hoffmann et al., 2006). The production wells are located in the Coyote Wash fault zone (see Fig. 1). The wells have high levels of dissolved CO2, which has been problematic

for TEP, requiring that most of their groundwater pro- duction wells be replaced with stainless-steel well strings due to

Fig. 1. Site map of study area at Springerville–St. John. Water levels are indicated by blue

contours. Red dots show wells sampled. Red box illustrates extent of numerical model.

(For interpretation of the references to color in this legend, the reader is referred

to the web version of the article.)

corrosion problems (TEP personal communication). Previous work- ers have concluded that the elevated dissolved CO2 observed in the TEP groundwater wells and at Salado Springs is sourced from the underlying CO2 dome, and that this CO2 is migrating upwards primarily along the Coyote Wash Fault zone (Gilﬁllan et al., 2011; Moore et al., 2005). Furthermore, the extensive travertine deposits in the area are attributed to CO2–rock–water interaction mov- ing carbonate-saturated groundwater to the surface along faults (Crumpler et al., 1994). The noble gas and stable isotope geochem- istry of the CO2 source and dissolved gases in the groundwater suggests a mantle-derived magmatic source for the CO2 that is likely associated with the Springerville volcanic ﬁeld (Gilﬁllan et al., 2008, 2011). Allis et al. (2005) conducted multi-phase reactive-transport simulations of the reservoir, leakage along the fault zone, and travertine deposition. They were able to explain the timing and total volume of travertine deposition by assuming an inﬂux of CO2 into the base of the fault (from below) of ∼63 kTon/year CO2. Their simulations showed a transient overpressurization of the reservoir, leading to a rise in the water table elevation and broad area of water and CO2 discharge. Less is known about the migration of brine from the reservoir layers into the shallow aquifer. The highest measured salinity in the aquifer (2150 ppm, Moore et al., 2005) is elevated compared to regional groundwater but much less saline than deeper water (at least 4210 ppm and probably much higher due to anhydrite lay- ers). Compared to the C aquifer chemistry at locations distal to the Springerville–St. Johns CO2 system, the salinity observed here is sig- niﬁcantly higher. For example, 200 km northwest of Springerville, the C aquifer has a chloride levels <115 mg/l, with arsenic lev- els below the maximum contaminant level as deﬁned by the U.S. EPA (Hoffmann et al., 2006). It is unclear whether the slightly ele- vated

salinity is related to the rising CO2 from depth. Simulations developed by Keating et al. (2010) suggested that upward ﬂow

164 E. Keating et al. / International Journal of Greenhouse Gas Control 25 (2014) 162–172

Fig. 2. Durov diagram showing variation in major ion chemistry, pH, and TDS. Blue circle is Salado Springs, open green circles are production wells, red diamond is a reservoir sample

(Moore et al., 2005). (For interpretation of the references to color in this legend, the reader is referred to the web version of the article.)

of free-phase CO2 from depth along the Coyote Wash fault zone could not displace the fresh water/saline water boundary upwards more than ∼100 m. It is interesting that according to water quality records maintained by Montgomery and Associates (2011), there has been little to no increase in salinity in the TEP production wells over time, despite relatively large amounts of extraction. There are several conclusions to be drawn from the observations in and above the Springerville–St. Johns reservoir. One, although the integrity of the caprock is sufﬁcient to preserve a signiﬁcant reservoir of CO2 (and high helium, ∼1%, Rauzi, 2003) within the deeper sedimentary units, high levels of deep-source dissolved CO2 in the shallow aquifer demonstrate a pathway for CO2 migration. A likely pathway to explain the CO2 occurrence in shallow water is the nearby relatively broad, high-permeability fault zone (Coyote Wash Fault zone). These characteristics have combined to make the Springerville–St. Johns CO2 dome a natural analog for leakage from a geologic carbon sequestration storage site. A variety of questions can be asked including (1) can the CO2 be identiﬁed in the shallow groundwater; (2) can CO2 migration along the fault zone explain the levels of dissolved CO2 seen in the aquifer; (3) is the CO2 migrating alone or with brine; and (4) has the leakage from the CO2 storage site affected the shallow groundwater chemistry?

3. Water chemistry

3.1. Methods

Groundwater samples were collected from 8 wells that Tucson Electric Power uses to produce cooling water for the Springerville Generating Station (Fig. 1). These 8 wells were in service at the time of sampling. Two travertine spring mounds and one shallow alluvial well were also sampled at Salado Springs (Fig. 1). The production wells operate at high ﬂow rates so purging of the well bore was unnecessary, and these wells are closed systems at the well-head location preventing degassing and oxygenation during pumping to the surface. These wells are also constructed of stainless steel to minimize corrosion from CO2-charged groundwater. Water was collected from taps located either on the wellhead or on piping leading directly from the wellhead. Prior to sample collection, water

was allowed to ﬂow slowly (to minimize degassing) until tempera- ture, pH and conductivity stabilized, usually within a few minutes. At Salado Springs, grab samples were collected from the two spring mounds. The alluvial well was under slight artesian pressure so it produced water from a spigot that was allowed to discharge until ﬁeld parameters stabilized. Samples for major cation and trace metals analyses were ﬁeld ﬁltered (0.45 m) and acidiﬁed with trace metal grade nitric acid. Samples for anions, alkalinity and sta- ble isotope analyses were collected unﬁltered with no headspace. Carbon stable isotope samples were collected into gas tight bottles to prevent degassing and fractionation prior to analysis. All samples were processed and analyzed at the Los Alamos National Laboratory Earth and Environmental Sciences Geochem- istry and Geomaterials Research Laboratories (GGRL). Major cations and trace metals were measured by ICP-OES and ICP-MS, and anions were determined by Ion Chromatography. Stable isotope ratios of dissolved inorganic carbon, and oxygen and hydrogen in water were determined by continuous ﬂow isotope ratio mass spectrom- etry (CF-IRMS) on a GV Instruments Isoprime mass spectrometer attached to a Multiﬂow peripheral instrument. Results are reported using standard delta notation in per mil (‰). ı13C is reported versus PDB, and ı18O and ıD are reported versus SMOW. Standard devi- ations on ı13C, ı18O and ıD measurements based on in-house and international standards were 0.13, 0.12 and 0.5‰ respectively.

4. Results

Table 1 summarizes the temperature, ﬁeld pH and major ion ele- ment chemistry for the well and spring samples. For the purpose of this study, we consider major elements to be Ca2+, Mg2+, Na+, K+, Cl−, and the anion complexes: HCO3−, SO42−. The log pCO2 and saturation indices for calcite and dolomite were determined using PHREEQC (Parkhurst and Appelo, 1999) using the Wateq4 database. Calcite was near saturation (|log SI| < 0.5) in 5 samples, undersatu- rated in six. Dolomite was supersaturated (log SI > 0) in about half of the samples. All other mineral phases were undersaturated in the samples and are not reported. Selected minor and trace elements are reported in Table 2. Only elements that were detected or are

E. Keating et al. / International Journal of Greenhouse Gas Control 25 (2014) 162–172 165

Table 1 Field

parameters and major element chemistry (mg/l) for TEP wells and Salado Springs.

Sample ID T (◦C) pH Ca Cl HCO3 K Mg Na SO4 Charge balance (%) Log pCO2a Log SIa

cal dol

P-7a 16.6 6.9 135 146 458 10 39 131 259 −3.7 −1.4 −0.1 0.5 P-8a

25.3 6.3 165 190 522 17 45 182 345 −2.1 −0.9 −0.6 −0.4 P-9a

25.3 6.4 315 491 754 27 54 382 710 −1.1 −0.8 −0.1 0.3 P-10a

24.8 6.3 184 213 563 17 45 205 379 −1.9 −0.9 −0.6 −0.4 P-16

22.7 6.9 139 138 467 12 41 137 252 −0.6 −1.4 −0.0 0.8 P-18

22.4 6.7 301 414 710 26 58 338 692 −3.5 −1.0 0.1 0.8 P-19

24.0 6.6 188 215 559 20 50 211 402 −1.0 −1.1 −0.2 0.4 P-20

24.9 6.4 163 175 516 16 42 166 323 −2.4 −1.0 −0.6 −0.3 Salado

1 (spring) 17.1 6.1 322 473 744 25 61 359 710 −3.8 −0.7 −0.8 −0.9 Salado 2 (spring) 17.3 6.2 309 449 758 26 59 380 707 −3.0 −0.8 −0.6 −0.6 Salado

3 (Well) 14.8 6.2 307 449 769 27 60 387 688 −1.2 −0.8 −0.7 −0.7

a Log pCO2 and log saturation index (SI) calculated using PHREEQC (Parkhurst and Appelo, 1999). cal = calcite; dol = dolomite.

Table 2 Selected

minor and trace element chemistry (mg/l) for TEP wells and Salado Springs.

Sample ID As B Ba Br Cs F Fe Li Mn Rb SiO2 Sr U

P-7a 0.003 0.30 0.023 nd 0.02 1.8 nd 0.18 0.01 0.044 15 1 nd P-8a

0.012 0.53 0.022 nd

0.02 2.3 0.3 0.24 0.01 0.055 16 2 nd P-9a 0.042 0.77 0.019 nd 0.04 1.9 0.8 0.57 0.04 0.107 16 3 nd P-10a

0.018 0.46 0.023 nd 0.02 1.9 0.2 0.26 0.02 0.062 17 2 nd P-16

0.011 0.35 0.025 nd 0.02 1.9 0.3 0.18 0.01 0.049 15 1 nd P-18

0.035 0.71 0.016 0.1 0.03 2.6 0.8 0.47 0.03 0.100 15 3 nd P-19

0.016 0.79 0.022 0.2 0.02 1.8 0.4 0.25 0.02 0.069 16 2 nd P-20

0.014 0.41 0.024 nd 0.02 2.2 0.2 0.23 0.02 0.061 15 1 nd Salado

1

(spring) 0.016 0.75 0.017 0.2 0.03 2.3 nd 0.55 0.03 0.100 17 3 nd Salado 2 (spring) 0.009 0.88 0.015 0.2 0.03 2.2 nd 0.50 0.02 0.098 17 3 nd Salado

3 (Well) 0.021 0.81 0.015 0.3 0.03 2.1 15 0.53 0.15 0.102 18 3 nd

of particular interest for groundwater quality (e.g., uranium) are tabulated.

A Durov diagram, Fig. 2, shows the variation in water chemistry at the site. For comparison, we included a sample from the deep reservoir collected by Moore et al. (2005). This ﬁgure illustrates the geochemical distinction between the shallow aquifer and the deep reservoir. The deep reservoir is enriched with respect to sodium, has a higher pH and TDS. Scatterplots show a consistent relation- ship between all major elements in the shallow aquifer. Fig. 3 presents representative relationships between HCO3−, Cl−and SO42−. The pattern that emerges is a linear array with a high con- centration end-member that includes wells P-9a and P-18 and all the Salado Spring locations. The other wells create a lower concen- tration linear array. There is a gap between the high concentration end member and lower concentration array that is ∼200 mg/l in terms of chloride. These patterns also hold for Li and Rb, as shown by the strong correlation with chloride (Fig. 4). Arsenic, the only trace element that was found elevated above EPA MCLs (10 ppb, or 0.010 mg/l), shows a slightly different behavior. The lin- ear array exists for the P-well samples, but note that the Salado Spring samples fall consistently off this trend (Fig. 2, As vs. Cl) where arsenic concentrations are signiﬁcantly lower than P-9a and P-18.

Results of the stable isotope analyses are reported in Table 3. Sta- ble isotopes O and H are consistent with regional meteoric water values (Bills et al., 1990; Kendall and Coplen, 2001). The ı13C of dis- solved inorganic carbon (DIC) measured in this study range from −1.6 to 0.28‰ (PDB). Gilﬁllan et al. (2011) found similar values in these wells, but also were able to access P-14 that returned a value of 0.9‰. The ı13CDIC from deep CO2 wells in the gas resource to the east of the production well ﬁeld fall in a narrow range between −3.8 and −3.6‰ (PDB) (Gilﬁllan et al., 2008, 2009). Although previ- ous workers did not see a strong signal in the carbon isotopes, we see

a correlation between P-well ı13CDIC and the major and trace element chemistry. Essentially, the highest carbon isotopes values (Table 3; P-9a, P-18 and P-14) correlate with the high concentration

Fig. 3. Selected trace elements (As, B, Li and Rb) versus chloride content. The binary mixing

trend seen in the major elements is preserved in the trace elements. Note that the

As concentrations at Salado Springs fall off this trend, with lower concentrations than

predicted by the mixing line.

166 E. Keating et al. / International Journal of Greenhouse Gas Control 25 (2014) 162–172

Fig. 4. Selected trace elements (As, B, Li and Rb) versus chloride content. The binary mixing trend seen in the major elements is preserved in the trace elements. Note that the

As

concentrations at Salado Springs fall off this trend, with lower concentrations than predicted by the mixing line.

(more saline) end-member. Similar to arsenic, the Salado Springs samples fall off this trend with a lower ı13CDIC.

5. Discussion

The major ion and trace element trends observed in the ground- water and Salado Springs can be explained by simple mixing of two sources (Figs. 3 and 4). It appears that a CO2-rich end member with higher salinity and elevated trace element concentrations is mix- ing with more dilute groundwater in the aquifer. This is supported by the co-variation of chloride with bicarbonate, sulfate, and most other major and trace elements. However, a simple mix of two end-members is not adequate to explain the carbon isotope trends observed. As Gilﬁllan et al. (2011) noted, the carbon isotopic composition of the groundwa- ter is difﬁcult to explain based on measurements of the deep CO2 source; whereas, gas migration and dissolution could explain the

Table 3 Stable

isotopes of O, H and C for TEP wells, Salado Springs, and deep gas wells.

Sample ID ı18O ‰ (SMOW) ıD ‰ (SMOW) ı13CDIC ‰ (PDB)

P-7a −10.96 −79.55 −1.62 P-8a

−10.65 −78.17 −1.78 P-9a

−10.76 −78.03 0.28 P-10a

−10.89 −78.27 −1.63 P-16

−10.81 −78.91 −1.87 P-18

−10.92 −79.10 −0.08 P-19

−10.91 −78.11 −1.29 P-20

−10.63 −75.88 −1.87 Salado

1 (spring) −10.59 −78.57 −0.33 Salado

2 (spring) −10.72 −78.80 −0.25 Salado

3 (Well) −10.78 −79.41 −1.09 P-14a nr nr 0.90 Deep

CO2 wellsa nr nr −3.6 to −3.8

a From Gilﬁllan et al. (2011).

noble gas compositional trends. Fig. 5 shows that the ı13CDIC and bicarbonate content of the proposed deep CO2 source do not fall on the mixing trend observed in the overlying aquifer. Also, the Salado Spring samples seem to fall off this trend. A proposed mechanism to

Fig. 5. Carbon stable isotope composition of the dissolved inorganic carbon (ı13CDIC) versus

total dissolved inorganic carbon (DIC) – for the TEP wells, Salado Springs, and the

deep CO2 wells (Gilﬁllan et al., 2011). The TEP wells form an apparent binary mixture

between high ı13CDIC and high DIC- end-member and low ı13CDIC, more dilute

groundwater. Salado Springs fall off this trend. A Rayleigh degassing models for

groundwater at 30 ◦C can explain evolution of the deep CO2 source to the high concentration

end-member seen in the groundwater (modiﬁed from Newell et al., 2008;

Gilﬁllan et al., 2009). The carbon isotope fractionation between the degassing CO2 and the total DIC pool for the model is calculated based on the speciation of DIC (CO2(aq) + HCO3−) using the approach of Deines et al. (1974) and Gilﬁllan et al. (2009): ε13CDIC-CO2 (g) = x(ε13CH2CO3 (aq)−CO2(g)) + (1 − x)(ε13CHCO3 (aq)−CO2(g)), where x is the fraction

of H2CO3 and (1 − x) is the fraction of HCO3−. The speciation is calculated using

PHREEQC (Parkhurst and Appelo, 1999) and the reported water chemistry for deep

CO2 well 11–21 (Moore et al., 2005). Since degassing during sampling may have

impacted the pH for this well, we adjusted the pH and total DIC by adding CO2 to reach calcite saturation (using PHREEQC), and present a second degassing model based

on this starting chemistry.

E. Keating et al. / International Journal of Greenhouse Gas Control 25 (2014) 162–172 167

explain the data is open system Rayleigh degassing (Fig. 5) (modi- ﬁed from Newell et al., 2008; Gilﬁllan et al., 2009). The fractionation between the DIC pool in the groundwater and the degassing CO2 is calculated assuming a temperature of 30 ◦C and considering the speciation of the DIC pool (aqueous CO2 and bicarbonate) (Deines et al., 1974; Gilﬁllan et al., 2009). The starting DIC and chemistry for speciation calculations using PHREEQC are based on State Well 11–21 reported in Moore et al. (2005). Although the charge bal- ance for this water chemistry is acceptable, the reported pH may be altered from in situ conditions due to degassing during samp- ling. Since the speciation is sensitive to the pH, we also present a second degassing model using modiﬁed starting water chemistry (Fig. 5). The DIC content and speciation and pH of the groundwater are modiﬁed in PHREEQC by adding CO2 until calcite saturation is reached (SI adjusted from 0.4 to 0.0). These simple models shows that 40–50% degassing can evolve the CO2 reservoir source ﬂuid to the ı13CDIC and total DIC concentration observed for the high con- centration end-member. This model is path independent in that this degassing could occur during ﬂuid ascent from source to ground- water, or from the shallow groundwater after ascent and mixing. This model may also explain the two Salado Spring samples col- lected

at travertine spring mounds, with a slightly smaller degree of degassing. The outlier Salado Spring sample is the shallow alluvial well sample; this sample could be experiencing mixing with allu- vial groundwater. Alternatively, microbial activity at Salado Springs could be modifying the carbon isotopic composition through pro- cesses such as dissolved organic matter oxidation or microbially mediated calcite precipitation. However, given the large inorganic carbon pool in the spring waters, these processes would not sig- niﬁcantly alter the dissolved inorganic carbon concentration. It is important to note that this Rayleigh degassing model cannot be modiﬁed to explain the dilute end-member of the groundwater trend. We suggest that a CO2-rich ﬂuid is migrating and evolv- ing due to degassing and that this evolved ﬂuid then mixes with groundwater to create the binary trend. The primary importance of dissolved CO2 transport rather than free-phase transport is sig- niﬁcant.

Also of interest is the source of salinity and trace elements creat- ing the observed mixing trends. The values of chloride and arsenic (491 and 0.042 mg/l) are higher than background values (<2 g/l) observed in waters to the north (Hoffmann et al., 2006). Emplace- ment of the Springerville volcanic ﬁeld could have introduced trace elements such as As, Li and B due to magmatic devolatilization and local hydrothermal circulation (Condit et al., 1989; Guilbert and Park, 1986). However, this cannot explain the addition of signif- icant amounts of chloride, sulfate and other major elements (e.g., calcium, sodium). A more plausible source of brine and trace metals is the Supai Group encountered along ﬂow paths below the aquifer. High arsenic values up to 1000 ppb in Verde Valley springs south- west of Springerville, are attributed to sources within the Supai Group (Foust et al., 2004; Johnson et al., 2012). Evaporite horizons, including thick anhydrite beds in the Big A Butte Member show evi- dence of CO2-brine-rock interaction (e.g., formation of secondary dawsonite) (Moore et al., 2005) are potential sources of these trace elements in the Supai Group. Modern salars are known to concen- trate trace elements such as As, B, and Li and result in associated groundwater enrichment in these elements (Concha et al., 2010; Risacher and Fritz, 1991). A notable exception to the binary mixing trends is the concen- tration of arsenic at Salado Springs (Fig. 4). Based on other major and trace elements, Salado Springs appears quite similar to wells P-9a and P-18 that deﬁne the high concentration end member. It is beyond the scope to fully investigate what processes are respon- sible for lowering the arsenic concentration, while not impacting other

trace elements. However, microbial processes coupled to iron and sulfur redox chemistry in the springs could be responsible for

Fig. 6. Rate of extraction for TEP production wells from 1981 to 2010 (total for all wells).

lowering arsenic concentration through the precipitation of iron oxides (Kirk et al., 2004). This interpretation is also consistent with the ı13CDIC of the springs falling off the trend. The carbon stable iso- topic values could have been lowered by microbial processes that preferentially use the lighter carbon isotope (Sharp, 2006). In summary, evidence from the water chemistry measurements is consistent with a conceptual model of brackish or saline water and CO2 moving up along the fault zone. The water transports salinity and trace metals from the layers within the Supai Group. Mixing with shallow dilute groundwater reduces concentrations; the resulting variability can be explained by a conservative, non- reactive mixing. Waters sampled in and near Salado Springs show evidence of some reactions, perhaps biogeochemical. Stable iso- topes suggest the CO2 has undergone signiﬁcant degassing. This suggests the CO2 moved upward in the dissolved phase, and degassed as pressures decreases or as the water mixed with shallow groundwater. The common theme of upward ﬂow of CO2-enriched water explains the transport of CO2, major and trace elements in this system. This does not preclude the possibility that CO2-induced changes in shallow water quality might be occurring here, rather, such changes are probably much more subtle than the changes caused by mixing with saline water.

6. Fault permeability

We used hydrologic data and groundwater ﬂow simulations to explore the hydrologic characteristics of the fault system. This system is unique in that production wells are completed near the fault, and complete extraction and drawdown data are avail- able for the wellﬁeld. A substantial amount of water has been produced from these wells (see Fig. 6) with relatively minor draw- down (34 m at most; see Fig. 1). An interesting feature of the drawdown response is that monitoring wells on the upthrown side of the fault (to the north) have experienced no measurable drawdown (see Fig. 1). We developed a series of 3-D ground- water ﬂow simulations with the goal of developing and testing hypotheses about the hydrologic characteristics of the fault zones, and attempting to match the drawdown data (to ﬁrst order). There is considerable uncertainty about the properties of the rock layers and the fault zone; to our knowledge there are no site- speciﬁc hydraulic testing results available. Therefore, for each conceptual model of the fault system, we consider the hydrologic properties of the rocks to be uncertain parameters that can be optimized to best match measured drawdowns. If two concep- tual models match the drawdowns equally well, they are equally likely.

We assume the same hydrostratigraphy for all the models, since this is fairly well known (Fig. 7). A simpliﬁed hydrostratigraphic

168 E. Keating et al. / International Journal of Greenhouse Gas Control 25 (2014) 162–172

Table 4 Sum

of squared errors for simulated drawdowns, using three conceptual models of the fault zone.

Well types Flow-neutral fault zone Distinct fault zone, shallow

only

Distinct

fault zone, full penetration

Production wellﬁeld (n = 15) 1087 5358 598 Monitoring

wells (n = 6) 133 10 6

model accounting for the Precambrian basement, four units of the Supai formation, the Glorieta Sandstone, and the San Andreas Lime- stone was developed based on interpretations in Moore et al. (2005) and Rauzi (1999). This 3-D model reproduces the measured offset in all the rock layers along the Coyote Wash fault. All layers above the San

Adreas, including the Chinle Formation, were combined into a single undifferentiated unit. Boundary conditions and annual extraction rates were deﬁned in the same way for all models. The lateral model extent (20 km × 20 km) was selected to be large enough to encompass all the wells with transient head data available yet not so large as to be computationally intractable. Unfortunately none of the lateral

boundaries coincide with groundwater divides. The closest groundwater divide, according to the Arizona Water Atlas (2011), is much too distant to the south to be a practical model bound- ary for this study. To account for the natural inﬂow and outﬂow of shallow groundwater along model boundaries (true ﬂux rates are unknown), constant pressure boundary conditions were assigned along all four boundaries in the shallow (above Supai) layers. Pressures were deﬁned in accordance with regional water level data derived from the Arizona Department of Water Resources database (2011). The upper model boundary was considered to be no-ﬂow, due to the conﬁning nature of the Chinle formation. Transient sim- ulations were conducted, applying annual extraction rates to each of 14 TEP production wells over a 30-year period (see Fig. 6). Ini- tial pressures were speciﬁed to be hydrostatic. At the end of each transient simulation, total drawdown over the entire period was compared to measured drawdown at each well with available data (21 wells total). Simulations were single-phase (water only), exe- cuted using the code FEHM (Zyvoloski, 1997) developed at Los Alamos National Laboratory. We tested three conceptual models for the fault zone: a ‘ﬂow neutral’ model, and two hydrologically distinct fault zones: one, penetrating only to the base of the deepest production well (∼1525 m), the other, penetrating the entire thickness of the model. In each case, PEST was given the freedom to vary permeabilities and

Fig. 7. Three-dimensional grid for the production simulations. Blue lines show location

of monitoring wells; red lines show location of production wells. (For inter- pretation

of the references to color in this legend, the reader is referred to the web version

of the article.)

Table 5 Rock

properties and sensitivities speciﬁc storage 10−4.0/m intrinsic permeability (log

10 m2).

Unit Permeability Sensitivity

Basement −17. – Supai

formation

Amos

Wash −14. –

Big Butte −14.4 – Ft.

Apache −12.1 – Corduroy

−14.7 9.077109E−03 Glorieta

Sandstonea −16. 1.769560E−02 San

Andreas Limestone −11.9 0.850563 Upper

units −11.0 0.793787 Fault

zone (upper, z

> 1525 m)

−11.3

4.91668

Fault

zone (lower, z

< 1525 m)

−12.5 8.54851

a Note: Due to the very low sensitivity of this parameter, the ‘estimated’ value (−16.) is not meaningful.

speciﬁc storage of all rock layers and (if present) fault zones within a wide range (10−11–10−16 m2 and 10−2–10−6/m, respectively). Results are summarized in Tables 4 and 5. Of the three con- ceptualizations of the fault zone, the model that considered the fault zone distinct, deep, and permeable outperformed the other two conceptualizations. The ‘ﬂow neutral’ fault zone conceptual model tended to overpredict drawdowns in the monitoring wells on the upthrown side of the fault (Table 3). Parameter sensitivities (Table 5) show that the highest sensitivity is to the permeabili- ties of the fault zone, particularly the deep portion. Because of the high sensitivities, these permeability estimates are associated with a higher level conﬁdence. Collectively, these results suggest that to reproduce measured drawdowns in both the wellﬁeld and the monitoring network to the northeast requires a hydrologically distinct fault zone that extends much deeper than the wellﬁeld itself. The match is not perfect (see Fig. 8), as is to be expected

Fig. 8. Water level drawdown over 29 years in uppermost model layer (m). Contours show

simulation results, ﬁlled squares indicate measurements.

E. Keating et al. / International Journal of Greenhouse Gas Control 25 (2014) 162–172 169

Fig. 9. 2-D Simulations of bouyancy-driven CO2 migration in the presence of a permeable fault zone: permeability variation.

when the hydrostratigraphy is generalized; however the simulated drawdowns are similar to those measured. By examining inﬂows along the constant pressure boundaries we were able to conﬁrm that the water extraction does increase inﬂow along the closest (southwestern) model boundary. How- ever, the increase in inﬂow is only about 15% of the extraction; the remaining 85% is derived from storage. Because of this rela- tively small effect, we do not believe these results are particularly sensitive to the pressure boundary conditions. This analysis demonstrates that the high permeability fault fea- ture that allowed signiﬁcant upﬂow of CO2 in the past, leading to large travertine deposits, is still a modern hydrologic pathway. The production wells are extracting water that comes, to some extent, from

depth. Unlike previous studies, this analysis constrains the depth and permeability of the fault zone.

7. The unlikely co-existence of a gas reservoir and a broad, leaky conduit

There are multiple

lines of evidence, including new analyses presented here, that the St. Johns–Springerville CO2 reservoir is penetrated by a broad, highly permeable fault zone. How, then, can an estimated 445 billion m3 of buoyant CO2 gas remain at depth? This question is

directly relevant to engineered CO2 sequestra- tion. Here we present multi-phase (CO2/water) ﬂow simulations to

investigate the stability of a shallow gas reservoir in the presence of a wide leaky fault. To simplify the system we extracted a 2-D cross-section from the 3-D model described above. The cross-section is 100 m thick, and represents approximately 1/500th the total reservoir vol- ume (see Fig. 9). We initialized the simulations with a gas-phase CO2 reservoir in multiple layers, according to the conceptual and numerical models presented by Allis et al. (2005). The pore space in the gas reservoir is 60% free-phase CO2, 40% water. The gas is allowed to dissolve in the water up to the pressure/temperature- dependent solubility limit. The total CO2 mass of the initial reservoir is ∼7 × 108 kg, approximately 30% in the dissolved phase. Pressure is initially hydrostatic, with a slight horizontal gradient in the shal- low

aquifer (.028). This gradient is within the range of measured gradients in the shallow aquifer at Springerville (approximately 0–0.03; Fig. 1). Figs. 10 and 11 illustrates the rapid changes that occur in the ﬁrst 300 days of simulation. Due to buoyancy, the gas begins to move upward along the broad high-permeability fault zone. Quickly, the fault zone is ﬁlled with a dissolved CO2 plume, which eventually moves away from the fault in the shallow aquifer. After a short time, however, the rate of gas depletion of the reservoir sharply declines.

The rate of gas depletion, shown in Fig. 12, becomes approxi- mately a constant value of 6 × 105 kg/year. To place this in context

Fig. 10. 2-D Simulations of bouyancy-driven CO2 migration in the presence of a permeable fault zone. (a) 0 days, (b) 8 days, (c) 98 days, and (d) 300 days.

170 E. Keating et al. / International Journal of Greenhouse Gas Control 25 (2014) 162–172

Fig. 11. 2-D Simulations of buoyancy-driven CO2 migration in the presence of a permeable fault zone. Dissolved CO2 mass fraction (a) 0 days, (b) 8 days, (c) 98 days, and (d) 300 days.

of the reservoir at Springerville, this is signiﬁcantly smaller than the leakage rate inferred by Allis et al. (2005) during the period of high CO2 degassing and travertine deposition. Our ﬁnding is thus consis- tent with the view that contemporary degassing of the Springerville reservoir is much less than during an earlier period in its history when travertines were being formed (Moore et al., 2005). Stevens et al. (2005) report CO2 in place at Springerville as 16 Tcf (trillion cubic feet at 21 ◦C and atmospheric pres- sure) or 445 billion m3. At representative in situ conditions (T = 25 ◦C, P = 4.5 MPa), CO2 density is approximately 60 times higher (112 kg m−3 vs. atmospheric 1.8 kg m−3) and the volume of gas in place is therefore ∼7.4 billion m3. The mass of CO2 is reported at 830 Mt. If the leak rate indicated by the 100-m thick cross-sectional model is scaled to a 50 km along-strike length (6 × 105 * 500 = 3 × 108 kg/year) then it is possible to address the following question: could a natural gas reservoir sustain these rates of depletion for long periods of time? If there were no CO2 replen- ishment at depth from the mantle, it would take approximately

Fig. 12. Total free-phase CO2 mass in the reservoir (excluding the fault zone).

2800 years to deplete a reservoir the size of St. John’s Dome at Springerville.

An interesting aspect of these results is that a large plume of dissolved CO2 is maintained within the fault zone long after the primary upward pulse of buoyancy-driven gas phase. The geom- etry of this plume is consistent with the presence of high levels of dissolved CO2 in the TEP wells. The slow leak rate of CO2 gas (Fig. 12) is probably adequate to maintain this dissolved CO2 plume indeﬁnitely.

Although we used some of the features of the Springerville site, such as the hydrostratigraphy and fault geometry, to inform the simulations, these simulations are not meant to replicate trans- port of CO2 speciﬁcally at the site, historically or modern. Rather, they are designed to address generic questions of reservoir sta- bility in the presence of faults. For example one key difference between these simulations and the evolution of the Springerville site is that the fault did not abruptly open after the CO2 was emplaced. Another key difference is that CO2 discharged to the surface along the fault zone at Springerville, whereas in these sim- ulations CO2 can only move laterally in the shallow groundwater aquifer.

Although this simulation does not mimic actual processes of reservoir development at Springerville, it does provide insights into the mechanisms allowing buoyant CO2 gas to stay in place despite the presence of vertical leaky conduits. Once the gas within and near the leaky feature escapes to the shallow aquifer, the remaining gas reservoir leaks at a very slow rate. Future studies would be use- ful for determining the impact of speciﬁc factors such as caprock thickness, reservoir thickness, and radial distance to the fault for gas reservoir stability.

8. Conclusions

The CO2 reservoir at Springerville–St Johns provides a natural analog for understanding mechanisms of long-term CO2 storage and leakage. Additionally, new insights can be gained regarding

E. Keating et al. / International Journal of Greenhouse Gas Control 25 (2014) 162–172 171

the impacts of CO2 leakage on shallow groundwater systems. The primary mechanism of long-term CO2 storage at this site, and many other natural gas reservoirs has been described previously as ‘sol- ubility trapping’ (Gilﬁllan et al., 2009). At this site, there appears to be a substantial amount of CO2 stored in both the dissolved and free (gas) phase (Rauzi, 1999). The simulations here show that buoyant free phase CO2 can remain at depth despite the presence of a large scale, high permeability leakage path. This is true even when caprock layers are nearly horizontal. This is because, under hydrostatic conditions, there is insufﬁcient driving force for the free phase CO2 to move laterally toward the fault before it can escape vertically.

Analysis of major, minor element geochemistry and stable iso- topes of carbon demonstrates that the leakage pathway (Coyote Wash Fault zone) has been a conduit for upward movement of CO2 dissolved in brackish or saline water. CO2 has degassed as the water moved upwards. Possible mechanisms causing upward movement of CO2-saturated water include: (1) overpressuriza- tion at depth due to mantle-derived CO2 inﬂow from basement penetrating faults (e.g., Crossey et al., 2009) or (2) groundwater overpressurization due to past glacial to interglacial transitions (e.g.,

Kampman et al., 2012). Importantly, this differs from the sce- nario presented by Keating et al. (2010), which considered free phase CO2 rising along the fault and displacing brine. Here, the primary transport mechanism is upward ﬂow of dissolved phase CO2.

The measured history of water extraction and consequence pressure declines at a wellﬁeld in the fault zone provided the oppor- tunity to characterize the hydrologic characteristics of the fault zone. Inverse modeling demonstrates that the offset of strata due to faulting is itself sufﬁcient explanation for the hydrologic compart- mentalization of TEP monitoring wells to the northeast. However, the fault zone must be broad, permeable, and deep to explain the measured drawdowns given the levels of production over the past few decades. This conclusion does not preclude the possibility that other conceptual models, not tested here, could be equally valid for the site. Elevated and co-varying concentrations of As, Li, B, Rb, Na, SO4 and Cl in the shallow aquifer demonstrate some degree of upward transport of brine or brackish water. The fairly linear relationship between trace elements and Cl suggest trace metals were not mobi- lized in situ due to CO2 intrusion. This result does not preclude the possibility that this might be the case at other sites. Carbon isotopes suggest that this water was likely saturated with CO2 and degassed as it moved upward along the fault. As noted by previous authors, the mass rate of CO2 migrating to the shallow aquifer is presum- ably much lower than 10’s to 100’s of thousands of years ago when extensive travertine deposits were created. Salado Springs are cur- rently undersaturated with respect to carbonates (SI, Table 1) and are precipitating travertine today only in small amounts as spring waters degas. Nevertheless, upward transport is still occurring. There are several possible implications of this study for engi- neered carbon sequestration and (possible) leakage. First, a breach in the caprock or activated fault may release some CO2 to the sur- face but buoyancy alone will not be sufﬁcient to release CO2 distant from the leakage pathway. Second, the release of CO2 to the shal- low aquifer will not necessarily mobilize trace metals in the shallow aquifer. Our measurements at this site do not support such a mecha- nism. Third, as previous authors have suggested, mobilization and trapping of CO2 in the dissolved phase will be the most impor- tant process (Gilﬁllan et al., 2009). At Springerville–St Johns dome, present or past pressure conditions forced brackish or saline water upward along the fault, and with it dissolved CO2. A signiﬁcant amount of degassing occurred as the CO2 rose. These processes are

quite different than the conceptual model of buoyancy-driven gas-phase ﬂow in a leakage scenario.

This site could be useful for further ﬁeld and simulation stud- ies aimed at better understanding CO2 ﬂow and transport in the subsurface. For example, alternative conceptual models of the site could be tested more rigorously using integrated multi-phase ﬂow/reactive transport simulations and comparison of model results to stable isotope and trace element data. New insights might be gained by sampling a larger number of wells in the area, and ana- lyzing for other trace gases such as helium. Finally, issues related to heterogeneity of fault zones, and impact on CO2 transport could be explored with focused studies of this site.

References

Aiuppa, A., Federico, C., Allard, P., Gurrieri, S., Valenza, M., 2005. Trace metal model- ing of groundwater–gas–rock interactions in a volcanic aquifer: Mount Vesuvius, Southern Italy. Chem. Geol. 216, 289–311. Allis, R.G., Moore, J., White, S., 2005. Reactive Multiphase Behavior of CO2 in Saline Aquifers Beneath the Colorado Plateau. Report to DOE, 23 pp. Arizona Department of Water Resources, 2011. Groundwater Site Inventory. https://gisweb.azwater.gov/waterresourcedata/GWSI.aspx (2011 retrieval). Arizona Department of Water Resources, 2011, Water Atlas. http://www.azwater. gov/AzDWR/StatewidePlanning/wateratlas/ (2001 retrieval). Bills, D.J., Hjalmarson, H.W., Robertson, F.N., 1990. Estimates of Ground-Water Flow Components for Lyman Lake, Apache County, Arizona: U.S. Geological Survey Water-Resources Investigations

Report 89-4151., pp. 55. Birkholzer, Jens, T., Zhou, Q., Tsang, C.-F., 2009. Large-scale impact of CO2 storage in deep saline aquifers: a sensitivity study on pressure response in stratiﬁed systems. Int. J. Greenhouse Gas Control 3, 181–194. Carey, J.W., Williams, M.R., Hakala, A., Fessenden, J.E., Keating, E.H., 2009. Exper- imental study of sediment-CO2 reactions with application to changes in groundwater quality due to leakage of sequestered CO2. In: American Geophys- ical Union, Fall Meeting, Abstract, H13A-0913. Concha, G., Broberg, K., Grander, M., Cardozo, A., Palm, B., Vahter, M., 2010. High- level exposure to lithium, boron, cesium, and arsenic via drinking water in the Andes of Northern Argentina. Environ. Sci. Technol. 44, 6875–6880. Condit, C.D., Crumpler, L.S., Aubele, J.C., 1989. Patterns of volcanism along the south- ern margin of the Colorado Plateau: the Springerville ﬁeld. J. Geophys. Res. 94, 7975–7986. Crossey, L.J., Karlstrom, K.E., Springer, A.E., Newell, D., Hilton, D.R., Fischer, T., 2009. Degassing of mantle-derived CO2 and He from springs in the southern Colorado Plateau region—neotectonic connections and implications for groundwater sys- tems. GSA Bull., http://dx.doi.org/10.1130/B26394.1. Crumpler, L.S., Aubele, J.C., Condit, C.D., 1994. Volcanoes and Neotectonic Character- istics of the Springerville Volcanic Field, Arizona. New Mexico Geological Society Guidebook, 45th Field Conference, Mogollon Slope, West-Central New Mexico and East-Central Arizona., pp. 147–164. Deines, P., Langmuir, D., Harmon, R., 1974. Stable carbon isotope ratios and the exis- tance of a gas phase in the evolution of carbonate ground waters. Geochim. Cosmochim. Acta 38, 1147–1164. Foust Jr., R.D., Mohapatra, P., Compton-O’Brien, A.M., Reifel, J., 2004. Groundwater arsenic in the Verde Valley in central Arizona, USA. Appl. Geochem. 19, 251– 255. Gilﬁllan, S.M.V., Ballentine, C.J., Holland, G., Blagburn, D., Sherwood Lollar, B., Stevens, S., Schoell, M., Cassidy, M., 2008. The noble gas geochemistry of natural CO2 gas reservoirs from the Colorado Plateau and Rocky Mountain provinces, USA. Geochim. Cosmochim. Acta 72, 1174–1198. Gilﬁllan, S.M.V., Sherwood Lollar, B., Holland, G., Blagburn, D., Stevens, S., Schoell, M., Cassidy, M., Ding, Z., Zhou, Z., Lacrampe-Couloume, G., Ballentine, C.J., 2009. Solubility trapping in formation water as dominant CO2 sink in natural gas ﬁelds. Nature 458, 614–617. Gilﬁllan, S.M.V., Wilkinson, M., Haszeldine, R.S., Shipton, Z.K., Nelson, S.T., Poreda, R.J., 2011. He and Ne as tracers of natural CO2 migration up a fault from a deep reservoir. Int. J. Greenhouse Gas Control 5, 1507–1516. Guilbert, J.M., Park, C.F., 1986. The Geology of Ore Deposits. W.H. Freeman and Company, New York, pp. 985. Hoffmann, J.P., Bills, D.J., Phillips, J.V., Halford, K.J., 2006. Geologic, Hydrologic, and Chemical Data from the C Aquifer Near Leupp, Arizona: U.S. Geological Survey Scientiﬁc Investigations Report 2005-5280., pp. 42. Hovorka, S.D., Benson, S.M., Doughty, C., Freifeld, B.M., Sakurai, S., Daley, T.M., Kharaka, Y.K., Holtz, M.H., Trautz, R.C., Nance, H.S., Myer, L.R., Knauss, K.G., 2006. Measuring permanence of CO2 storage in saline formations: the Frio experiment. Environ. Geosci. 13 (2), 105–121. Hovorka, S.D., Meckel, T.A., Trevino, R.H., Lu, J., Nicot, J.-P., Choi, J.-W., Freeman, D., Cook, P., Daley, T.M., Ajo-Franklin, J.B., Freifeild, B.M., Doughty, C., Carrigan, C.R., Brecque, D.L., Kharaka, Y.K., Thordsen, J.J., Phelps, T.J., Yang, C., Romanak, K.D., Zhang, T., Holt, R.M., Lindler, J.S., Butsch, R.J., 2011. Monitoring a large volume CO2 injection: year two results from SECARB project at Denbury’s Cranﬁeld, Mississippi, USA. Energy Proc. 4, 3478–3485. Johnson, R.H., DeWitt, E., Wirt, L., Manning, A.H., Hunt, A.G., 2012. Using geochem- istry to identify the source of groundwater to Montezuma Well, a natural spring in Central Arizona, USA: Part 2. Environ. Earth Sci. 67, 1837–1853.

172 E. Keating et al. / International Journal of Greenhouse Gas Control 25 (2014) 162–172

Kampman, N., Burnside, N.M., Shipton, Z.K., Chapman, H.J., Nicholl, J.A., Ellam, R.M., Bickle, M.J., 2012. Pulses of carbon dioxide emissions from intracrustal faults following climatic warming. Nat. Geosci. 5, 352–358. Keating, E.H., Newell, D.L., Viswanathan, H., Carey, J.W., Zyvoloski, G., Pawar, R., 2012. CO2/brine transport into shallow aquifers along fault zones. Environ. Sci. Technol. 47 (1), 290–297. Keating, E.H.<Et-Al>, 2010. The impact of CO2 on shallow groundwater chemistry: observations at a natural analog site and implications for carbon sequestration. Environ. Earth Sci., http://dx.doi.org/10.1007/s12665-009-0192-4. Keating, E.H., Hakala, J.A., Viswanathan, H., Carey, J.W., Pawar, R., Guthrie, G.D., Fessenden-Rahn, J., 2013. CO2 leakage scenarios: simulations informed by obser- vations at a natural analog site. Appl. Geochem. 30, 136–147. Kendall, C., Coplen, T.B., 2001. Distribution of oxygen-18 and deuterium in river water across the United States. Hydrol. Proc. 15, 1363–1393. Kharaka, Y.K., Thordsen, J.J., Kakouros, E., Ambats, G., Herkelrath, W.N., Beers, S.R., Birkholzer, J.T., Apps, J.A., Spycher, N.F., Zheng, L., Trautz, R.C., Rauch, H.W., Gul- lickson,

K.S., 2010. Changes in the chemistry of shallow groundwater related to the 2008 injection of CO2 at the ZERT ﬁeld site, Bozeman, Montana. Environ. Earth

Sci. 60, 273–284. Kirk, M.F., Holm, T.R., Park, J., Jin, Q., Sanford, R.A., Fouke, B.W., Bethke, C.M., 2004. Bacterial sulfate reduction limits natural arsenic contamination in groundwater. Geology 32 (11), 953–956. Korbol, R., Kaddour, A., 1995. Sleipner-vest CO2 disposal – injection of removed CO2 into the Utsira formation, 1995. Energy Convers. Manage. 36 (6–9), 509–512. Little, M.G., Jackson, R.B., 2010. Deep CO2 geosequestration on overlying freshwater aquifers. Environ. Sci. Technol. 44, 9225–9232. Montgomery and Associates, 2011. Hydrogeologic Monitoring Program 2010-201, Springerville Generating Station Area Apache County, Arizona; prepared for Tucson Electric Power. Moore, J., Adams, M., Allis, R., Lutz, S., Rauzi, S., 2005. Mineralogical and geochemical consequences of the long-term presence of CO2 in natural reservoirs: an example from the Springerville–St. Johns Field, Arizona, and New Mexico, U.S.A. Chem. Geol. 217, 365–385. Newell, D.L., Jessup, M.J., Cottle, J.M., Hilton, D., Sharp, Z., Fischer, T., 2008. Aque- ous and isotope geochemistry of mineral springs along the southern margin

of the Tibetan plateau: implications for ﬂuid sources and regional degassing of CO2. Geochem. Geophys. Geosyst. 9, Q08014, http://dx.doi.org/10.1029/ 02008GC002021. Oldenburg, Curtis, M., Rinaldi, A.P., 2011. Buoyancy effects on upward brine dis- placement caused by CO2 injection. Trans. Porous Med. 87, 525–540. Parkhurst, D.L., Appelo, C.A.J., 1999. User’s Guide to PHREEQC (Version 2) – A Com- puter Program for Speciation, Batch-Reaction, One-Dimensional Transport, and Inverse Geochemical Calculations: USGS Water-Resources Investigations Report 99-4259., pp. 310. Peter, A., et al., 2012. Investigation of the geochemical impact of CO2 on shallow groundwater: design and implementation of a CO2 injection test in Northeast Germany. Environ. Earth Sci. 7 (2), 335–349. Preisig, M., Prevost, J.H., 2011. Coupled multi-phase thermo-poromechanical effects. Case study: CO2 injection at In Salah, Algeria. Int. J. Greenhouse Gas Control 5 (4),

1055–1064. Rauzi, S., 2003. Review of Helium Production and Potential in Arizona, Open-File Report

03-05. Arizona Geological Survey, 30 pp. Rauzi, S., 1999. Carbon Dioxide in the St. Johns–Springerville Area, Apache County, Arizona.

Open-File Report 99-2. Arizona Geological Survey, 24 pp. Risacher, F., Fritz, B., 1991. Quaternary geochemical evolution of the salars of Uyuni and Coipasa, Central Altiplano, Bolivia. Chem. Geol. 90, 211–231. Sharp, Z.D., 2006. Principles of Stable Isotope Geochemistry. Pearson/Prentice Hall, Upper Saddle River, NJ, pp. 344. Smyth, R.C., Hovorka, S.D., Lu, J., Romanak, K.D., Partin, J.W.C., Yang, W.C., 2009. Assessing risk to fresh water resources from long term CO2 injection-laboratory and ﬁeld studies. Energy Proc. 1, 1957–1964. Stevens, S.H., Fox, C., White, T., Melzer, S., 2005. Natural CO2 Analogs for Carbon Sequestration. Final Report for DOE. Trautz, R.C., Pugh, J.D., Varadharajan, C., Zheng, L., Bianchi, M., Nico, P.S., Spycher, N.F., Newell, D.L., Esposito, R.A., Wu, Y., Dafﬂon, B., Hubbard, S.S., Birkholzer, J.T., 2013. Effect of dissolved CO2 on a shallow groundwater system: a controlled release ﬁeld experiment. Environ. Sci. Technol. 47, 298–305. Zyvoloski, G.A., 1997. Summary of the Models and Methods for the {FEHM} Appli- cation – A Finite-Element Heat- and Mass-Transfer Code; Los Alamos National Laboratory Report, 72 pp.