CO2 Leakage-Induced Arsenic Contamination
and the Associated Health Risk Assessment
Chan Yeong Kim
The Graduate School
Yonsei University
Department of Earth System Sciences
CO2 Leakage-Induced Arsenic Contamination
and the Associated Health Risk Assessment
A Masters Thesis
Submitted to the Department of Earth System Sciences
and the Graduate School of Yonsei University
in partial fulfillment of the
requirements for the degree of
Master of Geological Science
Chan Yeong Kim
December 2017
This certifies that the masters thesis
of Chan Yeong Kim is approved.
___________________________
Thesis Supervisor: Weon Shik Han
___________________________
Nam Chil Woo
___________________________
Eungyu Park
The Graduate School
Yonsei University
December 2017
감사의 글
학위논문을 마무리 하며 그 동안 많은 도움과 격려를 주신 소중한
분들께 감사의 글을 올립니다.
먼저 부족한 저를 한국에 오신 그 순간부터 석사과정을 잘 마무리 할
수 있기까지 많은 가르침과 깨달음을 주시고 지도해주신 우리 한원식
교수님께 진심으로 감사를 드립니다. 저를 대학원에 올 수 있도록
이끌어 주시고, 또 흔쾌히 심사를 맡아 주신 우남칠 교수님께도
감사드립니다. 늘 피와 살이 되는 조언을 아낌없이 해주시고 바쁘신
와중에 멀리서 오셔서 심사를 맡아 주신 박은규 교수님께도
감사드립니다. 학부 때부터 지구시스템과학과에서 암석학, 원격 탐사,
수리-지화학, 구조지질학 등 다양한 분야에서의 가르침을 주신 권성택,
원중선, 송윤구, 이기현, 권상훈, 이용재, 홍태경, 김진욱 교수님께
감사드립니다.
석사과정을 하는 동안 저와 모든 시간들을 함께한 길택이형,
기돈이형 사랑하고, 연구실에서 늘 저를 응원해준 종길이, 태권이, 소담
누나 감사드립니다. 지구시스템학과의 동기들, 선후배님들께도 감사의
인사를 전합니다. 모두 다 뜻하는 바를 이루리라 굳게 믿습니다.
끝으로 저의 힘, 사랑하는 부모님과 찬우 그리고 가족들에게
감사드립니다.
2017 년 12 월
i
Table of Contents
List of Figures ⅲ
List of Table vii
ABSTRACT ix
1. Introduction 1
2. Behavior of Arsenic in Subsurface Environment 5
3. Work Flow: Numerical to Probabilistic Quantification 7
4. Numerical Approach 10
5. Conceptual Model 13
5.1. Model Description 13
5.2. Physical and Chemical Parameter 15
5.3. Probabilistic Health Risk Approach 24
6. Model Scenarios 28
7. Result and Discussion 30
7.1. Base case 30
7.1.1. Migration of Leaked CO2 plume within the Sallow Potable Aquifer 30
7.1.2. Induced Geochemical Reactions 33
7.1.3. Health Risk Assessment for Carcinogenic Effect 41
ii
7.2. Sensitivity Studies 45
7.2.1. Effect of Sorption Intensity (Kd) (Cases 2-4) 45
7.2.2. Effect of CO2 Leakage Intensity (QL) (Cases 5-7) 50
7.2.3. Effect of Horizontal Permeability (kh) (Cases 8-10) 55
8. Conclusion 61
9. References 64
국문요약 78
iii
List of Figures
Figure 1. Eh-pH diagram for arsenic species under H-O-As system at 1 bar and 25℃ 6
Figure 2. Flow chart explaining the sequence of both numerical simulation and probabilistic
health risk assessment. Numerical simulation composed of multi-phase and multi-species
reactive transport modeling required both physical and chemical parameters within the
shallow aquifer. Subsequently, the output of numerical simulation was utilized for
calculating probabilistic health risk for carcinogenic effect 9
Figure 3. Conceptual model representing the shallow potable aquifer with a length and
thickness of 200 m and 40 m, respectively. Blue arrows represent ambient groundwater
flow with the rate of 2 cm/day. The grid-blocks with the purple color stand for constant
pressure boundaries; left and right boundary is 3.2 MPa and 2.8 MPa respectively. Both
upper and lower boundaries are assumed to be the no-flow boundary. CO2 leakage point
(red) is located at 50 m away from the left boundary. In addition, the municipal well (green)
located at 50 m away from the right boundary penetrates 30 m deep from the upper seal
while the pumping activity was only conducted through the screened section 14
iv
Figure 4. Saturation of leaked CO2 at different times (120, 240, 360, 480, and 600 days).
Red, black, and yellow circles with labels indicate the locations of mass center for CO2
plume. Red circles in Fig 4d and 4e stand for the evolution of the mass center in multiple
times; yellow circles in Fig. 4e express the mass center from 600 to 1,080 days 32
Figure 5. Distribution of pH, carbonate minerals (Calcite and Magnesite), and associated
dissolved species (HCO3-, Ca2+, and Mg2+) after 365 days 35
Figure 6. Distribution of silicate minerals (K-feldspar, Illite, Chlorite, and Kaolinite) and
associated dissolved species (SiO2(aq), K+, and AlO2) after 365 days 37
Figure 7. Distribution of altered permeability (%) after 365days. The changes in
permeability was predicted from the following equation (𝐤′(%) =𝐤𝐜𝐡𝐚𝐧𝐠𝐞𝐝−𝒌𝒊𝒏𝒊𝒕𝒊𝒂𝒍
𝒌𝒊𝒏𝒊𝒕𝒊𝒂𝒍×𝟏𝟎𝟎)
because its variation was too small. Here, kinitial and kchanged indicates the initial permeability
and altered permeability at designated time, respectively 38
Figure 8. Distribution of Arsenopyrite dissolution and associated dissolved species (∑As,
Fe2+, HS-, O2(aq)) after 365 days 40
v
Figure 9. (a) Changes in pH, total arsenic (∑As), and arsenite (H3AsO30) concentrations
monitored at the municipal well. The Maximum Contaminant Level (MCL) of arsenic
species is 1.33⨯10-7 mol/L. Gaseous CO2 saturation (Sg) and pH for 10 years was also
plotted at the magnified window. Two stages were characterized based on the behavior of
CO2 plume. (b) Frequency of predicted health risk accounting for both ∑As and H3AsO30,
and cumulative density functions representing probabilistic health risk. The bin size of both
histograms was designated to be 10-4, and thus, the number of bins was 10 and 20,
respectively. Finally, the threshold of risk or the risk level of concern was 10-4 (red dotted
line) 42
Figure 10. (a) ∑As profiles with different Kd from 25, 50, to 100 L/kg-1 (Cases 2-4) and (b)
probabilistic health risk plotted as cumulative density functions. Red dotted-line represents
the risk level of concern (10-4) 47
Figure 11. (a), (b), and (c) Saturation of leaked CO2 at different leakage rate after 1 year
(365 days) (Cases 5-7). (d), (e), and (f) Distribution of ∑As concentration induced by
leaked CO2 at different leakage intensity (QL) after 20 years (Case 5-7) 51
Figure 12. (a) ∑As profiles with different leakage intensity (QL) from 0.01, 0.02, to 0.03
vi
kg/s (Cases 5-7) and (b) probabilistic health risk for Case 5-7 plotted as cumulative density
functions (CDF). Red dotted-line represents the risk level of concern (10-4) 54
Figure 13. (a), (b), and (c) Saturation of leaked CO2 at different horizontal permeability (kh)
after 1 year (365 days) (Cases 8-10). (d), (e), and (f) Distribution of ∑As Concentration
induced by leaked CO2 at different horizontal permeability (kh) after 20 years (Cases 8-10)
58
Figure 14. (a) ∑As profiles with different horizontal permeability (kh) from 0.2× 10-13,
1.0× 10-13, to 5.0× 10-13 m2 (Cases 8-10) and (b) probabilistic health risk for Case 8-10
plotted as cumulative density functions (CDF). Red dotted-line represents the risk level of
concern (10-4) 59
vii
List of Table
Table 1. Input parameters for shallow groundwater aquifer materials. Relative permeability
and capillary pressure follow equations developed by Van Genuchten (1980) 16
Table 2. Mineral volume fractions for shallow groundwater aquifer materials. A total of 7
primary and 5 secondary minerals were selected 18
Table 3. The 16 primary species and 39 secondary species are selected after the batch-
reaction simulation 19
Table 4. Kinetic data for selected primary and secondary minerals. Only calcite was treated
similar to the equilibrium reaction, and the rest minerals were kinetically treated with
consideration of neutral, acid, and base mechanisms 21
Table 5. Parameters for health risk assessment. Each parameter is randomly sampled within
designated value range and distributions to represent individual characteristics of potential
victims 26
viii
Table 6. Different parameters for sensitivity studies. Three groups except Case 1 (Base-
case) are designed for sorption rate (Kd) (Case 2-4), gaseous CO2 leakage rate (QL) (Case
5-7), horizontal permeability (kh) (Case 8-10), respectively 28
ix
ABSTRACT
Leakage of stored CO2 from a designated deep reservoir could contaminate overlying
shallow potable aquifers by dissolution of arsenic-bearing minerals. To ellucidate CO2
leakage-induced arsenic contamination, 2D multi-species reactive transport models were
developed and simulated CO2 leakage processes in the shallow groundwater aquifer.
Throughout a series of numerical simulations, it was revealed that the movement of leaked
CO2 was primarily governed by local flow fields within the shallow potable aquifer. The
induced low pH plume caused dissolution of aquifer minerals and sequentially increased
permeabilities of the aquifer; especially, the most drastic increase in permeability appeared
at the rear margin of CO2 plume where two different types of groundwater mixed. The
distribution of ∑As plume was similar to one for the Arsenopyrite dissolution. The break-
through curve of ∑As monitored at the municipal well was utilized to quantify the human
health risk. In addition, the sensitivity studies were conducted with different sorption rate
of arsenic species, CO2 leakage rate, and horizontal permeability in the aquifer. In
conclusion, the human health risk was influenced by the shape of ∑As plume, which was,
in turn, affected by the characteristics of CO2 plume behavior such as horizontal
permeability and CO2 leakage rate.
Key words: CO2 sequestration, Reactive transport modeling, Risk assessment, Arsenic.
1
1 Introduction
Carbon capture and storage is considered to be one of the mitigating strategies for
reducing CO2 emissions to the atmosphere (Metz et al., 2005; NETL, 2013; Reichle et al.,
1999). Among various carbon capture and storage technologies, CO2 can be injected into
geologically stable formations, which typically have large storage capacities and are capped
by low-permeability sealing formations. However, during CO2 injection activity, injection-
induced pressure builds up within the storage formation (Birkholzer et al., 2009; Guyant et
al., 2016; Strandli and Benson, 2013). The resulting pressure gradient could induce vertical
movement of injected supercritical CO2, which is less dense than the ambient brine. If any
unwanted pathways exist within the sealing formation, CO2 is able to migrate to the shallow
aquifer through these pathways while experiencing a phase change from supercritical to
gaseous CO2 (Bachu, 2008; Keating et al., 2009; Oldenburg et al., 2012; Pruess, 2008).
Keating et al. (2009) reveled two types of the CO2 leakage near unwanted pathways such
as fault. One is that CO2 rises with deep water as aqueous phase or solely gaseous phase
along a fault and forms point source such as CO2-rich spring and the other one is that CO2
rises along faults as aqueous or gaseous phase and diffuses into aquifer which located above
reservoir. Especially in this study, the leakage of gaseous CO2 through unexpected pathway
into a shallow confined aquifer was considered. Leaked gaseous CO2 dissolves into the
potable groundwater and develops a low pH plume (Apps et al., 2011; Kharaka et al., 2006),
which induces the secondary contamination within the aquifer by enhancing the mobility
of toxic heavy metals (Carroll et al., 2014; Keating et al., 2013a; Wang and Jaffe, 2004;
2
Zheng et al., 2015).
Released toxic heavy metals are able to migrate with the ambient groundwater if a
hydraulic gradient is developed by any active municipal wells nearby. If the contaminated
groundwater produced from the municipal well is directly distributed without proper
treatment (disinfection or chlorination processes) to residents who use it daily for the
purpose of drinking, bathing, cleaning, or other household uses, these residents can be
exposed to adverse carcinogenic health risks. The World Health Organization’s (WHO)
International Agency for Research on Cancer classified various dissolved heavy metals
(e.g., arsenic and lead) and other radioactive metals (e.g., uranium and cesium) as toxic
substances hazardous to human health (WHO, 2012). Among these carcinogenic heavy
metals, this study focused on arsenic (Matschullat, 2000). According to the US Agency for
Toxic Substances and Disease Registry, who ranked hazardous substances based on their
occurrence, toxicity, and potential for human exposure, arsenic was ranked the first in their
Substance Priority List in 2013 and 2015 (ATSDR). Moreover, arsenic contamination and
its detrimental impacts have been reported by various countries such as China (Rodríguez-
Lado et al., 2013), Bangladesh (Chowdhury et al., 2000), Vietnam (Meharg and Rahman,
2003), and India (Hebbar Annapoorna, 2016) in recent years (Nordstrom, 2002).
Mobilization of arsenic in a shallow groundwater aquifer due to CO2 leakage has been
investigated at one of the natural analog sites, Chimayó, New Mexico, where CO2-saturated
brackish-water was leaked into the shallow aquifer along the fault (Keating et al., 2009;
Keating et al., 2013a; Keating et al., 2013b; Viswanathan et al., 2012). At this site,
3
decreased pH and the resulting mobilization of trace metals, including arsenic, were
observed. Even if their adverse effects had been alleviated due to the high buffering
capacity of the local groundwater aquifer, Keating et al. (2009) reported significantly
elevated trace metal concentrations at a number of local wells due to the influx of brackish
waters. Later, both Keating et al. (2013a) and Viswanathan et al. (2012) integrated the field
dataset into a multi-phase reactive transport model to understand the behavior of arsenic,
since some wells in Chimayó exceeded the maximum contamination level (MCL). In
addition to studies targeting natural CO2 release sites, several experiments have been
conducted at field-scale CO2 injection sites to determine secondary contamination caused
by the injected CO2 (Kharaka et al., 2006; Kharaka et al., 2010; Trautz et al., 2013; Xu et
al., 2010). Decreased pH, increased HCO3- concentrations, and subsequent changes in
groundwater chemistry such as increased Fe2+, Mn2+, Mg2+, and Ca2+ concentrations were
observed at both Frio-I brine pilot injection (Kharaka et al., 2006; Xu et al., 2010) and
ZERT field tests (Apps et al., 2011; Kharaka et al., 2010).
Addition to field-oriented research, several studies have focused on the
implementation of numerical simulations to evaluate geochemical behaviors associated
with arsenic contamination. For example, Zhang et al. (2016) and Xiao et al. (2017) utilized
the reactive transport simulation and investigated the complex chemical change induced by
CO2 leaked into a shallow aquifer. In contrast to these studies which presented detailed
geochemical interactions, Siirila et al. (2012) simplified the geochemical processes by
solving the advection-dispersion equation with linear sorption. Without full assessment of
4
geochemical reactions, they were able to account for the movement of toxic elements in
complex 3D heterogeneous systems as well as for groundwater contamination-induced
carcinogenic health risks with a probabilistic approach. Later, Navarre-Sitchler et al. (2013)
utilized PFLOTRAN and simulated the mobilization of lead in a complex heterogeneous
system by assuming that released gaseous CO2 instantaneously dissolved into the
groundwater aquifer. The proposed work in this study was built upon the framework of the
previous studies mentioned above. Certain approaches accounted for the detailed
geochemical behavior of toxic heavy metals (e.g., complexation, sorption, mineral
dissolution, and precipitation) in the groundwater aquifer, whereas other approaches relied
on either simplified chemical reactions (e.g., advection-dispersion equation) or multi-phase
fluid migration (e.g., gaseous CO2 leakage into the groundwater aquifer) while accounting
for carcinogenic health risks or complex 3D heterogeneous systems. Therefore, the goal of
this study was to integrate these two approaches and delineate the multi-phase behavior of
leaked gaseous CO2 to a shallow potable aquifer. Moreover, leaked CO2-induced
geochemical changes, such as evolution of water quality and mobility of toxic trace metals
(e.g., arsenic), were characterized by adapting the multi-species reactive transport model.
Finally, the simulated concentration of arsenic species observed from the assigned
municipal well was used to quantify the carcinogenic health risk for chronically exposed
humans.
5
2 Behavior of Arsenic in Subsurface Environment
Arsenic in natural water typically originates from arsenic-bearing minerals that
frequently possess sulfur, oxygen, and iron (Lu and Zhu, 2010). Generally, these naturally
occurring arsenic-bearing minerals include Arsenopyrite (FeAsS), Realgar (AsS), Enargite
(CuAsS4), Scorodite (FeAsO4∙2H2O), and Tennantite (Cu6[Cu4(Fe,Zn)2]As4S13)
(Bhattacharya et al., 1997; Bose and Sharma, 2002; Zhang et al., 2016). When these
minerals dissolve, various forms of arsenic species such as H3AsO40, H2AsO4
-, HAsO42-,
AsO43-, H3AsO3
0, and H2AsO3- can be released into groundwater. As seen in Figure 1,
arsenic species can be present in several valence states (-3, 0, +3, and +5), but in natural
groundwater it is mostly found in oxyanions of trivalent arsenite (As(III)) or pentavalent
arsenate (As(V)), depending on reducing or oxidizing conditions, respectively. Typically,
the mobility and toxicity of As(III) is considered to be much higher than that of As(V)
(Berman and Berman, 1980; Cutter, 1992; Jain and Ali, 2000). Under the reducing
condition, uncharged As(III) species, such as H3AsO30, is dominant below pH 9.2 (Figure
1). However, under the oxidizing condition, H2AsO4- is dominant at pH<6.9, while HAsO4
2-
is dominant at higher pH (pH>6.9). Moreover, H3AsO40 and AsO4
3- could exist under
extremely acidic and alkaline conditions, respectively. In this study, when injected CO2
leaked into the shallow potable groundwater aquifer, the dissolution of leaked CO2 induced
the acidification of the ambient groundwater to 4<pH<6 (Kharaka et al., 2006), where
H3AsO30 and H2AsO4
- were dominant under reducing and oxidizing conditions,
respectively (Lu and Zhu, 2010; Smedley and Kinniburgh, 2002). Here, the shallow aquifer
6
Figure 1. Eh-pH diagram for arsenic species under H-O-As system at 1 bar and 25℃.
7
was considered to be under the reducing condition, and thus, H3AsO30 was chosen as the
most dominant species.
3 Work Flow: Numerical to Probabilistic Quantification
The following study comprised two major parts, namely numerical prediction of CO2
transport from a leakage point to a municipal well (Numerical Simulation) and probabilistic
quantification of health risks to humans who have been chronically exposed to a certain
toxic heavy metal (Health Risk Assessment), in sequence (Figure 2). Processes for both
numerical simulation and probabilistic risk quantification were designed in three steps:
“Data”, “Process”, and “Result”. During the “Data” step, input parameters were chosen
selectively, and the sampling scheme was determined either deterministically or randomly.
Subsequently, a series of calculations using either numerical or probabilistic approaches
was conducted in the “Process” step. Finally, in the “Result” step, simulation outputs, such
as CO2 plume distribution, maximum concentrations of selected toxic heavy metals, and
human health risk, were analyzed. The link between “numerical simulation” and “health
risk assessment” was the profile of arsenic concentration monitored at the municipal well;
throughout complex spatial and temporal movements of the CO2 plume predicted from
numerical simulations, a time series of arsenic concentration was observed at the municipal
well, which was then used to assess carcinogenic human health risk.
8
Figure 2. Flow chart explaining the sequence of both numerical simulation and
probabilistic health risk assessment. Numerical simulation composed of multi-phase
and multi-species reactive transport modeling required both physical and chemical
paraemters within the shallow aquifer. Subsequently, the output of numerical simulation
was utilized for calculating probabilistic health risk for carcinogenic effct.
9
10
4 Numerical Approach
The multi-phase and -component reactive transport simulator, TOUGHREACT, was
used to simulate secondary contamination processes in a shallow confined aquifer induced
by CO2 leakage (Xu et al., 2006) in conjunction with the ECO2N module, which was used
to predict the fluid properties of H2O, NaCl, and gaseous CO2 (Pruess and Spycher, 2007).
For this study, the original thermodynamic database incorporated in TOUGHREACT was
not adequate for delineating the complex chemical reactions associated with arsenic species.
Therefore, the revised EQ3/6V7.2b database was specifically adopted to account for
chemical reactions of arsenic and related chemical species (Zheng et al., 2009). In detail,
the modified thermodynamic database adopted arsenite (H3AsO30) as a primary species and
incorporated associated aqueous complexes (e.g., H2AsO3-, H2AsO4
-, H3AsO4(aq),
HAsO2(aq), HAsO42-, and HAsS2(aq)), while accounting for their activity coefficients from
the extended Debye-Hückel equation (Helgeson et al., 1981). Moreover, solubility products
of arsenic-bearing minerals such as Arsenopyrite were included in this thermodynamic
database.
TOUGHREACT utilizes mathematical formulations to simulate chemical reactions
such as dissociation of aqueous complexes and dissolution/precipitation of minerals (André
et al., 2014; Xu, 2008). Aqueous reactions follow the assumption for the local chemical
equilibrium. With this premise, the concentration of secondary aqueous species can be
expressed as the product of primary species concentration as shown below:
11
Ci = 𝐾𝑖−1𝛾𝑖
−1 ∏ 𝐶𝑗
𝑣𝑖𝑗𝑁𝑐𝑗=1 𝛾
𝑗
𝑣𝑖𝑗 (Eq. 1)
Where, 𝐶𝑖 is the molal concentration of the ith secondary species, 𝐶𝑗 is the molal
concentration of the jth primary species among the Nc number of primary species, 𝛾𝑖 and
𝛾𝑗 are the activity coefficients of secondary and primary species, respectively, 𝑣𝑖𝑗 is the
stoichiometric coefficient, and 𝐾𝑖 is the equilibrium constant.
Precipitation and dissolution of minerals were simulated kinetically by following the
rate law, which was coupled with the equation representing the kinetic rate constant (Eq.
2). In this equation, the kinetic rate constant is not only dependent on temperature but also
on the pH as shown below (André et al., 2014):
rn = ± {k25𝑛𝑢 exp [−
𝐸𝑎𝑛𝑢
𝑅(
1
𝑇−
1
298.15)] + k25
𝐻 exp [−𝐸𝑎
𝐻
𝑅(
1
𝑇−
1
298.15)] 𝑎𝐻
𝑛𝐻 +
k25𝑂𝐻 exp [−
𝐸𝑎𝑂𝐻
𝑅(
1
𝑇−
1
298.15)] 𝑎𝑂𝐻
𝑛𝑂𝐻} 𝐴𝑛|1 − Ω𝑛𝜃|
𝜂 (Eq. 2)
Where, Ea (J/mol) is the activation energy and k25 (mol/m2/s) is the rate constant
at 25°C with superscripts nu, H, and OH representing neutral, acid, and base mechanisms,
respectively. R (J/mol/K) is the gas constant, T (K) is temperature, α is the activity of
dissolved species under acid or base conditions, and n is the power term. An is the specific
reactive surface area (cm2/g), and Ωn is the kinetic mineral saturation ratio of the nth
mineral. Finally, both θ and η are assumed to be unity.
12
5 Conceptual Model
5.1 Model Description
The model designed for this study delineates the CO2 leakage process throughout
undetected and unexpected pathways in the sealing formation. In order to understand such
processes, the 2D cross-sectional potable aquifer model was designed as shown in Figure
3. The potable aquifer was assumed to be relatively deep (300 m with a reducing condition),
where the municipal or high-capacity wells produced a large amount of groundwater, which
was then distributed for residential purposes (Larsen et al., 2003; Renken et al., 2005). In
addition, the width and thickness of the potable aquifer was assigned as 200 m and 40 m,
respectively, with the size of grid-blocks as 2 m; the total number of grid-blocks was 2,000
(100 × 20). The size of grid represents the resolution of the model. Tighter resolution
(smaller grid) offers more accurate simulation, but requires substantially increased time for
its calculation. Hence, we tested three different grid models with different resolution and
selected the one with the optimal grid size and calculation time. Associated figures and
captions are included within the appendix.
Assuming that the aquifer is located at a depth of 300 m, the initial formation
pressure and temperature was assigned as 3 MPa and 25°C, respectively. Top and bottom
boundaries of the model were assigned as no-flow conditions assuming that the upper and
lower confining formations act as nearly impermeable sealing units. Lateral boundaries
(purple-colored grid-blocks) were assigned as the Dirichlet condition, where the left and
13
Fig
ure
3.
Co
nce
ptu
al m
odel
rep
rese
nti
ng t
he
shal
low
pota
ble
aquif
er w
ith a
len
gth
an
d t
hic
kn
ess
of
20
0 m
an
d 4
0 m
,
resp
ecti
vel
y. B
lue
arro
ws
repre
sent
ambie
nt
gro
undw
ater
flo
w w
ith the
rate
of
2 c
m/d
ay. T
he
gri
d-b
lock
s w
ith
the
pu
rple
colo
r st
and
for
const
ant
pre
ssure
boundar
ies;
lef
t an
d r
ight
boundar
y i
s 3.2
MP
a an
d 2
.8 M
Pa
resp
ecti
vel
y. B
oth
upper
and
lo
wer
bo
un
dar
ies
are
assu
med
to b
e th
e no
-flo
w b
oundar
y. C
O2 l
eakag
e poin
t (r
ed)
is l
oca
ted
at
50
m a
way
fro
m t
he
left
bo
un
dar
y. I
n a
dd
itio
n, th
e m
unic
ipal
wel
l (g
reen
) lo
cate
d a
t 50 m
aw
ay f
rom
th
e ri
gh
t b
ou
nd
ary p
enet
rate
s 3
0 m
dee
p
fro
m t
he
up
per
sea
l w
hil
e th
e pum
pin
g a
ctiv
ity w
as o
nly
conduct
ed t
hro
ugh t
he
scre
ened
sec
tio
n.
14
right pressure was 3.2 MPa and 2.8 MPa, respectively. The difference of pressure in lateral
boundaries addition to aquifer properties (kh=10-13 m2) results in 1.94 cm/day of the ambient
groundwater flow in the aquifer. The leakage point for gaseous CO2 was located at 50 m
distances from the left boundary where CO2 was leaked at a rate of 0.05 kg/s (Figure 3).
The CO2 leakage was maintained during 1 year. Furthermore, the municipal well with a
continuous pumping rate of 0.5 kg/s was located 100 m away from the CO2 leakage point
and penetrated a depth of 30 m from the upper confining seal. Finally, the simulation was
conducted for 100 years.
15
5.2 Physical and Chemical Parameters
The aquifer was assumed to be homogeneous with its horizontal permeability (kh) and
porosity (ϕ) to be 10-13 m2 and 0.2, respectively (Table 1); the anisotropy ratio (kv/kh) was
assumed to be 0.1. In addition, delineation of gaseous CO2 transport through the
groundwater-saturated aquifer required constitutive relations such as relative permeability
and capillary pressure. In this study, Van Genuchten’s functions were adapted, and the
relevant parameters are shown in Table 1 (Van Genuchten, 1980).
The mineralogical composition of the designated aquifer materials was assumed to be
sandstone, which is the typical hosting formation for geological CO2 sequestration
(Doughty et al., 2008; Finley et al., 2013) and natural analog CO2 leakage sites (Dockrill
and Shipton, 2010; Shipton et al., 2004). As an example, Kampman et al. (2014) analyzed
the mineralogical assemblage of Navajo Sandstone, which was considered to be the
primary sourcing aquifer for CO2-charged brine (Parry et al., 2007). Chemical analyses of
Navajo Sandstone fluids collected either from adjacent geysers or springs revealed elevated
concentrations of both arsenic and other toxic heavy metals (Carruthers, 2016; Han et al.,
2017; Kampman et al., 2009; Kampman et al., 2014). Moreover, Carruthers (2016)
provided summary of 27 literatures about laboratory batch experiments and CO2 injection
field experiments, all of which utilized sandstone as an aquifer and reservoir material.
Hence, the mineralogical composition in this study was based on the composition of Navajo
Sandstone, with the assumption that Arsenopyrite was the primary source of arsenic species
in the ambient groundwater.
16
Table 1. Input parameters for shallow groundwater aquifer materials. Relative
permeability and capillary pressure follow equations developed by Van Genuchten
(1980).
Aquifer Characteristic
Permeability Horizontal 10-13
Vertical 10-14
Porosity 0.2
Density (kg/m3) 2600
Tortuosity 0.3
Compressibility 10-8
Diffusivity (m2/s) 10-9
Relative Permeability (Van Genuchten, 1980)
𝑘𝑟𝑙 = √𝑆̅ {1 − (1 − [𝑆̅]1
𝑚)𝑚
}2
𝑆̅ =𝑆𝑙−𝑆𝑙𝑟
1−𝑆𝑙𝑟−𝑆𝑔𝑟
𝑘𝑟𝑔 = 1 − 𝑆̅ 2 1 − 𝑆̅2
𝑆𝑙𝑟: irreducible water saturation 0.20
𝑆𝑔𝑟: irreducible gas saturation 0.05
𝑚: exponent 0.457
Capillary Pressure (Van Genuchten, 1980)
pcap = −𝑃0 ([𝑆∗]−1
𝑚 − 1)1−𝑚
S∗ =𝑆𝑙−𝑆𝑙𝑟
1−𝑆𝑙𝑟
𝑆𝑙𝑟: irreducible water saturation 0.00
𝑚: exponent 0.457
𝑃0: strength coefficient 19.59 kPa
17
Total 7 primary minerals including Quartz, Kaolinite, Illite, K-feldspar, Arsenopyrite,
Calcite, and Magnesite were chosen, and another 5 secondary minerals (Chlorite, Dolomite,
Goethite, Oligoclase, and Smectite-Ca) were expected to be precipitated (Table 2). Mineral
volume fraction of each primary mineral was chosen from a similar quantity of Navajo
Sandstone with the addition of 1% Arsenopyrite; Quartz was predominant (81%), and
kaolinite accounted for the second-largest amount (16%). The Arsenopyrite occurs as
disseminated veins, grains, and nodules in the sandstone (Chen et al., 2015; Hills and
Thomas, 1945; Thornburg and Sahai, 2004). The volume fraction of Arsenopyrite is
spatially variable, but generally does not compose a substantial portion of material present,
mainly due to its form of existence. Hence, we assumed the volume fraction of Arsenopyrite
to be 1% within sandstone formations.
Due to the small proportion of carbonate minerals, fluid chemistry possessed less
buffering capacity against CO2 dissolution. With the chosen mineral composition, the batch
reaction was conducted to determine a list of primary species and their initial
concentrations. As a result, AlO2-, Ca2+, Cl-, Fe2+, H+, H2O, H3AsO3(aq), HCO3
-, HSO42-,
HS-, K+, Mg2+, Na+, O2(aq), SiO2(aq), and SO42- were selected as primary species, and
associated aqueous complexes were chosen as secondary species (Table 3).
As described by Keating et al. (2009) and Zheng et al. (2009), the concentration of
arsenic species in the aquifer media was significantly controlled by both adsorption and
desorption processes with not only clay minerals but also iron oxides. Due to this reason,
many researchers experimentally measured sorption values (Kd) for arsenic species under
18
Table 2. Mineral volume fractions for shallow groundwater aquifer materials. A total of
7 primary and 5 secondary minerals were selected.
Primary mineral Vol. Fraction
Quartz SiO2 8.10⨯10-1
Kaolinite Al2Si2O5(OH)4 1.58⨯10-1
Illite K0.6Mg0.25Al1.8(Al0.5Si3.5 O10) (OH)2 1.28⨯10-2
K-feldspar KAlSi3O8 9.43⨯10-3
Arsenopyrite FeAsS 9.00⨯10-3
Calcite CaCO3 3.38⨯10-4
Magnesite MgCO3 2.12⨯10-4
Secondary mineral Vol. Fraction
Chlorite Mg2.5Fe2.5Al2Si3O10(OH)8 0.00
Dolomite (CaMg(CO3)2) 0.00
Goethite FeO(OH) 0.00
Oligoclase CaNa4Al6Si14O40 0.00
Smectite-Ca Ca0.145Mg0.26Al1.77Si3.97O10(OH) 0.00
19
Table 3 The 16 primary species and 39 secondary species are selected after the batch-
reaction simulation.
Primary species (16)
AlO2- H+ HSO4
2- Na+
Ca2+ H2O HS- O2(aq)
Cl- H3AsO3(aq) K+ SiO2(aq)
Fe2+ HCO3- Mg2+ SO4
2-
Secondary species (39)
Al(OH)2+ CO2(aq) H2S(aq) MgHCO3
+
Al(OH)3(aq) CO32- H3AsO4(aq) MgSO4(aq)
Al3+ Fe3+ H3SiO4- NaCl(aq)
AlOH2+ FeCl+ HAlO2(aq) NaCO3-
CaCl2(aq) FeCl42- HAsO4(aq) NaHCO3(aq)
CaCO3(aq) FeCO3+ HAsO4
2- NaOH(aq)
CaHCO3+ FeHCO3
+ HAsS2(aq) NaSO4-
CaOH+ H2(aq) HS- OH-
CaSO4(aq) H2AsO3- HSO3
- SO2(aq)
CH4(aq) H2AsO4- MgCl+
20
various conditions (e.g., different minerals and pH) and revealed that the Kd of As(III) and
As(V) varied from 115 L/kg to over 10,000 L/kg within geologic media including
subsurface soil and aquifers (Hu et al., 2012; Sakata, 1987; Smedley and Kinniburgh, 2002).
In this study, a shallow sandstone aquifer was targeted, composed of over 90% Quartz and
Kaolinite (Table 2). Previous experimental measurements revealed that Kd for Quartz and
Kaolinite was measured at 2 and 19 L/kg, respectively (Smedley and Kinniburgh, 2002).
Therefore, among the broad range of sorption intensity value (Kd) of geologic media, the
value (10 L/kg) from the case with the most similar mineral volume fraction to the case
represented in Table 2, was randomly selected.
The multiple parameters required to address the kinetic rates for mineral reactions
following Eq. 2 are listed in Table 4. In addition, the calculation of reactive surface area for
minerals was followed by both Xu et al. (2010) and Sonnenthal et al. (2005), who assumed
that a mineral is a cubic array of truncated spheres, in which the radius of the sphere is
assumed to be 0.001 m. In this study, the surface roughness-based area predicted from the
spherical radius was reduced two orders to reasonably represent the reactive surface area.
Typically, chemical reactions only occur at selected sites on the mineral surface, and
furthermore, only a small fraction of mineral surface is involved in this reaction due to
grain coating and armoring. Therefore, the reactive surface areas of most silicate and
carbonate minerals were chosen to be approximately 10 cm2/g, similar to those chosen by
Knauss et al. (2005) and Zerai et al. (2006). Finally, reactive surface areas of clay minerals
such as Kaolinite, Illite, Chlorite, and Smectite were selected for larger values because of
21
Tab
le 4
Kin
etic
dat
a fo
r se
lect
ed p
rim
ary a
nd s
econdar
y m
iner
als.
Only
cal
cite
was
tre
ated
sim
ilar
to t
he
equ
ilib
riu
m
reac
tio
n, an
d t
he
rest
min
eral
s w
ere
kin
etic
ally
tre
ated
wit
h c
onsi
der
atio
n o
f neu
tral
, ac
id, an
d b
ase
mec
han
ism
s.
Min
eral
A
n
(cm
2/g
)
Neu
tral
Mec
han
ism
A
cid M
echan
ism
B
ase
Mec
han
ism
k2
5
(mo
l/m
2∙s
) E
a
(kJ/
mo
l)
k2
5
(mo
l/m
2∙s
) E
a
(kJ/
mo
l)
n
(H+)
k25
(mo
l/m
2∙s
) E
a (k
J/m
ol)
n
(H
+)
Qu
artz
9
.8
1.0
2⨯
10
-14
87.7
Kao
linit
e 1
.95⨯
10
5
6.9
1⨯
10
-14
22.2
4.8
9⨯
10
-12
65.9
0
.78
8.9
1⨯
10
-18
17
.9
-0.4
7
Illi
te
6.6
8⨯
10
5
1.6
6⨯
10
-13
35.0
1.0
5⨯
10
-14
23.6
0
.34
3.0
2⨯
10
-17
85
.9
-0.4
0
K-f
eld
spar
9
.8
3.8
9⨯
10
-13
38.0
8.7
1⨯
10
-11
51.7
0
.50
6.3
1⨯
10
-12
94
.1
-0.8
2
Ars
eno
pyri
te
12
.9
2.5
2⨯
10
-12
62.8
Cal
cite
E
quil
ibri
um
Rea
ctio
n
Mag
nes
ite
9.8
4.5
7⨯
10
-10
23.5
4.1
7⨯
10
-7
14.4
1
.00
Chlo
rite
9.8
3.0
2⨯
10
-13
88.0
7.7
6⨯
10
-12
88.0
0.5
0
Do
lom
ite
12
.9
2.5
2⨯
10
-12
52.2
2.3
4⨯
10
-7
43.5
0
.50
Go
eth
ite
9.8
2
.52⨯
10
-12
62.8
Oli
gic
alse
9
.8
1.4
4⨯
10
-12
69.8
2.1
3⨯
10
-10
65.0
0
.46
Sm
ecti
te-C
a 5
.64⨯
10
5
1.6
6⨯
10
-13
35.0
1.0
5⨯
10
-14
12.6
0
.34
3.0
2⨯
10
-17
58
.9
-0.4
0
22
their smaller grain sizes (Zheng et al., 2009). Porosity was associated with changes in
volume due to mineral dissolution or precipitation (Xu, 2008). The porosity in this study
was simulated by the following equation:
𝜙 = 1 − ∑ (𝑓𝑟𝑚 − 𝑓𝑟𝑢)𝑛𝑚𝑚=1 (Eq. 3)
Where nm is the number of minerals, frm and fru are the volume fractions of mth mineral
in the rock and non-reactive rock, respectively. Finally, permeability change was calculated
by using the porosity changes with cubic law (Steefel and Lasaga, 1994):
𝑘 = 𝑘𝑖 (𝜙
𝜙𝑖)
3 (Eq. 4)
Where k and 𝜙 are permeability and porosity, respectively, with subscript i
representing the initial value.
Even if the proposed approach was capable of simulating water-rock interaction, it
would still have some limitations. Firstly, in this study, the sorption effect of arsenic onto
the surface of clay minerals was simulated with a linear Kd approach instead of surface
complexation. Although the linear Kd approach had limitations in terms of delineating
chemical heterogeneities on both temporal and spatial scales, this approach was effective
for application to a large-scale simulation by reducing computation (Bethke and Brady,
2000; Davis et al., 1998; Valocchi, 1984). Secondly, Arsenopyrite is a solid solution of
Pyrite (FeS2), in which the ratio of arsenic and sulfur varies depending on their mole
fractions. However, in this study, for simplification in predicting the thermodynamic
23
properties, it was assumed that 1 mole fraction of arsenic replaced the sulfur (FeAsS).
Finally, it was assumed that Arsenopyrite oxidatively dissolves in the presence of common
geologic oxidants such as dissolved O2, NO3-, and Fe3+. Under acidic conditions, Fe3+
quickly oxidizes, and sequential dissolution of Arsenopyrite releases arsenic species
(H3AsO40 and H3AsO3
0) by following Eq. 5. Therefore, the stoichiometric reaction for
Arsenopyrite dissolution, which was used in this numerical simulation, involved the
reduction of Fe3+ to Fe2+ (Zheng et al., 2009).
FeAsS + 1.50 𝐻2𝑂 + 0.75 𝑂2(𝑎𝑞) + 𝐻+ = 𝐹𝑒2+ + 𝐻𝑆− + 𝐻3𝐴𝑠𝑂3(𝑎𝑞) (Eq. 5)
24
5.3 Probabilistic Health Risk Approach
Carcinogenic effects on humans who have been chronically exposed to arsenic species
through multiple pathways was probabilistically quantified based on individual exposure
rate and toxicity, which were suggested by the “Guidelines for Carcinogenic Risk
Assessment” of the US Environmental Protection Agency (EPA) (EPA, 2005), Siirila et al.
(2012), and the EPA Superfund Risk Assessment (Means, 1989). In general, three different
uptake pathways, namely ingestion, dermal sorption, and inhalation, were considered.
Among these, uptake via ingestion and dermal sorption was considered to be major
pathways, because humans are often exposed to risk by drinking dissolved toxic species in
tap water or showering. Uptake through inhalation was not considered in this study, as it is
unlikely that the concentration of vaporized trace metals was high enough to cause
carcinogenic effects indoors under normal conditions. Therefore, following Siirila et al.
(2012), only two exposure pathways (ingestion and dermal sorption) were considered for
quantifying exposure rate and toxicity of arsenic.
Risk = 𝑒1−(𝐴𝐷𝐷𝑖×𝐶𝑃𝐹𝑖) ≈ 𝐴𝐷𝐷𝑖×𝐶𝑃𝐹𝑖 (Eq. 6)
Arsenic toxicity was predicted from the product of cancer potency factor (CPF) (kg∙
day/mg) and average daily dose (ADD) (mg/kg∙day). Typically, the CPF is different at the
individual pathway even for the same toxic metal; in this study, the CPF for ingestion and
dermal sorption of arsenic were assumed to be 1.5 and 1.6 kg∙day/mg by following IRIS
25
(1991) and EPA (2004), respectively. Subsequently, ADD was used to assess the individual
exposure rate with the following equation:
ADD = 𝐶̅ [𝐼𝑁
𝐵𝑊]
𝐸𝐷×𝐸𝐹
𝐴𝑇 (Eq. 7)
Where [𝐼𝑁
𝐵𝑊] is the individual intake per body weight (L/kg/day), AT is the average
lifetime, which was assumed to be 25,550 days (70 years), and EF is the standard exposure
frequency during 1 year, which was assumed to be 350 (days/year) (Fahrner et al., 2012).
The most significant term in Eq. 7 is 𝐶̅ , which is the maximum average of arsenic
concentration (mg/L) monitored at the municipal well during the exposure duration (ED)
(years). In this study, the ED was assumed to be 30 years following EPA guideline (EPA,
1989). Accordingly, 𝐶̅, calculated from concentration profiles (C(t)) of arsenic species at
the municipal well in the numerical simulation, can be represented as follows:
𝐶̅ = 𝑚𝑎𝑥[∑ 𝐶𝑡+𝐸𝐷𝑡 (𝑡)]−∞
+∞ (Eq. 8)
Specific to ADD, exposure pathways through both ingestion and dermal sorption are
defined as follows:
ADD𝑖𝑛𝑔𝑒𝑠𝑡𝑖𝑜𝑛 = 𝐶̅ [𝐼𝑅
𝐵𝑊]
𝐸𝐷×𝐸𝐹
𝐴𝑇 (Eq. 9)
ADD𝑑𝑒𝑟𝑚𝑎𝑙 = 𝐶̅ [𝑆𝐴
𝐵𝑊]
𝐸𝐷×𝐸𝐹
𝐴𝑇Kp𝑓𝑠𝑘𝑖𝑛𝐸𝐷𝑠ℎ𝑜𝑤𝑒𝑟𝐶𝐹 (Eq. 10)
Where [𝑆𝐴
𝐵𝑊] is the skin surface area per body weight (m2/kg), Kp is the dermal
permeability coefficient for arsenic in water, designated as 1.010-5 m/hour, fskin is the
26
fraction of skin in contact with water (-), EDshower is exposure time of shower per day
(hours/day), and CF is the unit conversion factor (0.001 L/m3). Finally, the total risk is the
summation of individual risk representing each exposure pathway as shown below:
Total Risk = Risk𝑖𝑛𝑔𝑒𝑠𝑡𝑖𝑜𝑛 + Risk𝑑𝑒𝑟𝑚𝑎𝑙 = (ADD𝑖𝑛𝑔𝑒𝑠𝑡𝑖𝑜𝑛×CPF𝑖𝑛𝑔𝑒𝑠𝑡𝑖𝑜𝑛) +
(ADD𝑑𝑒𝑟𝑚𝑎𝑙×CPF𝑑𝑒𝑟𝑚𝑎𝑙) (Eq. 11)
In this study, the calculation of the total risk was repeated 100,000 times to account for
variability, which stands for probabilistic quantification in human health risk. Except for
𝐶̅ (the concentration of arsenic species obtained from the result of numerical simulation),
individual parameters considered in the quantification of total risk were randomly sampled
within each intrinsic distribution (Table 5) (Maxwell et al., 1998; Siirila et al., 2012).
Finally, calculated total health risks were plotted as cumulative density function, which
enables the estimation of the probability exceeding the risk level of concern (10-4) (EPA,
2001; McSlarrow, 1991).
27
Tab
le 5
. P
aram
eter
s fo
r hea
lth r
isk a
sses
smen
t. E
ach p
aram
eter
is
random
ly s
amp
led
wit
hin
des
ign
ated
val
ue
ran
ge
and
dis
trib
uti
ons
to r
epre
sent
indiv
idual
char
acte
rist
ics
of
pote
nti
al v
icti
ms.
Par
amet
er
Sym
bol
Unit
D
istr
ibu
tio
n
Val
ues
Ex
po
sure
dura
tio
n
ED
[y
r]
Const
ant
30
Ex
po
sure
fre
quen
cy
EF
[d
/yr]
C
onst
ant
35
0
Aver
agin
g t
ime
AT
[d
] C
onst
ant
25
,55
0
Inges
tio
n r
ate
per
un
it b
ody w
eight
IR/B
W
[L∙k
g/d
] L
ognorm
al
(3.3⨯
10
-2,
1.3⨯
10
-2)
Skin
su
rfac
e ar
ea p
er u
nit
body w
eight
SA
/BW
[m
2/k
g]
Lognorm
al
(2.7⨯
10
-2,
2.5⨯
10
-3)
Fra
ctio
n s
kin
in
co
nta
ct w
ith w
ater
f s
kin
[-]
Unif
orm
(4
.0⨯
10
-1,
9.0⨯
10
-1)
Sh
ow
er e
xp
osu
re d
ura
tion
ED
sho
wer
[h/d
] L
ognorm
al
(1.3⨯
10
-1,
9.0⨯
10
-2)
Unit
conver
sion f
acto
r C
F
[L/m
3]
Co
nst
ant
1.0⨯
10
-3
Can
cer
po
ten
cy f
acto
r C
PF
ing
esti
on
[kg∙d
ay/m
g]
Const
ant
1.5
CP
Fd
erm
al
1.6
28
6 Model Scenarios
Case 1 (base-case) was designed to delineate CO2 leakage processes associated water-
rock interactions, and secondary contamination caused by arsenic species (Table 6).
Subsequently, sensitivity studies were conducted with the different Kd, CO2 leakage rates,
and horizontal permeability of the aquifer (kh). First, the degree of sorption intensity was
evaluated by varying Kd from 25, 50, to 100 L/kg (Cases 2–4). Second, in Cases 5–7, the
effect of CO2 leakage rate was evaluated by varying its rate from 0.020, 0.025, to 0.030
kg/s. Different CO2 leakage rates could induce the development of CO2 plumes with
different sizes. For example, as the size of a CO2 plume increases, larger areas expect to
experience water-rock interactions and more dissolution of arsenic species. Consequently,
the municipal well captures dissolved arsenic species more when the size of the CO2 plume
is greater, which eventually increases the carcinogenic health risk on humans. Finally, in
Cases 8–10, kh varied (kh=0.210-13, 1.010-13, and 5.010-13 m2) while maintaining the
vertical permeability (1.010-14 m2); increased kh accelerates the horizontal velocity of
ambient groundwater while reducing buoyancy forces on the CO2 plume.
29
Tab
le 6
. D
iffe
ren
t p
aram
eter
s fo
r se
nsi
tivit
y s
tudie
s. T
hre
e gro
ups
exce
pt
Cas
e 1 (
Bas
e-c
ase)
are
des
ign
ed f
or
sorp
tio
n
rate
(K
d)
(Cas
e 2
-4),
gas
eous
CO
2 l
eakag
e ra
te (
QL)
(Cas
e 5
-7),
hori
zonta
l per
mea
bil
ity (
kh)
(Cas
e 8
-10
), r
esp
ecti
vel
y.
Cas
e n
ame
Sorp
tion r
ate
(Kd)
CO
2 L
eakag
e ra
te (
QL)
Ho
rizo
nta
l P
erm
eab
ilit
y (
kh)
Cas
e 1
(B
ase-
case
) 10
0.0
50 k
g/s
1
.0⨯
10
-13 m
2
Cas
e 2
25
0.0
50 k
g/s
1
.0⨯
10
-13 m
2
Cas
e 3
50
0.0
50 k
g/s
1
.0⨯
10
-13 m
2
Cas
e 4
100
0.0
50 k
g/s
1
.0⨯
10
-13 m
2
Cas
e 5
50
0.0
20 k
g/s
1
.0⨯
10
-13 m
2
Cas
e 6
50
0.0
25 k
g/s
1
.0⨯
10
-13 m
2
Cas
e 7
50
0.0
30 k
g/s
1
.0⨯
10
-13 m
2
Cas
e 8
50
0.0
50 k
g/s
0.2⨯
10
-13 m
2
Cas
e 9
50
0.0
50 k
g/s
1
.0⨯
10
-13 m
2
Cas
e 1
0
50
0.0
50 k
g/s
5.0⨯
10
-13 m
2
30
7 Results and Discussion
7.1 Base case
7.1.1 Migration of Leaked CO2 Plume within the Shallow Potable Aquifer
Figures 4a–4e represent the evolution of the leaked CO2 plume at designated times of
120, 240, 360, 480, and 600 days. The mass centers of the CO2 plume, shown as red, black,
and yellow circles, were calculated and plotted every 120 days; black circles represent the
present time of the mass center, and red and yellow circles represent past and future times,
respectively. From the leakage point, CO2 continuously leaked at a rate of 0.05 kg/s only
until 365 days. During this period, three-flow systems induced by ambient groundwater,
CO2 leakage, and pumping activity interacted with each other (Figure 4c), which developed
two mixing zones for geochemically different types of groundwater at both the front and
rear margins of the CO2 plume. The ambient groundwater flow (orange arrows) was
developed from the left to right boundaries at an approximate rate of 1.94 cm/day.
Additionally, the CO2 plume gradually expanded from the leakage point where CO2
saturation remained at 0.3; the rate of CO2 flux was approximately 2.110-5 kg/(s∙m2) (black
arrows) adjacent to the leakage point. At the rear margin of the CO2 plume, two chemically
different types of groundwater (ambient groundwater and CO2-dissolved groundwater)
flowed in opposite directions, inducing the development of a vigorous geochemical mixing
zone. Subsequently, the CO2 plume migrated together with the ambient groundwater until
it was captured by the municipal well (Figure 4c).
31
Figure 4. Saturation of leaked CO2 at different times (120, 240, 360, 480, and 600 days).
Red, black, and yellow circles with labels indicate the locations of mass center for CO2
plume. Red circles in Fig 4d and 4e stand for the evolution of the mass center in multiple
times; yellow circles in Fig. 4e express the mass center from 600 to 1,080 days.
32
After CO2 leakage had stopped at 365 days, the size of the CO2 plume gradually
decreased due to both dissolution to ambient groundwater and extraction from the
municipal well (Figures 4d and 4e); CO2 solubility predicted from Duan and Sun (2003)
was 0.83 mol/kg water in this aquifer (3 MPa and 25°C). After movable CO2 was captured
by the municipal well, residually trapped CO2 governed by irreducible CO2 saturation
(shown in Table 1) remained until complete dissolution to the ambient groundwater;
residually trapped CO2 eventually vanished approximately 5 years after CO2 leakage had
stopped.
The calculated mass center was located close to the plume center while the CO2 plume
migrated toward the municipal well. The migration rate of the CO2 plume estimated from
the mass center locations was approximately 12.5 cm/day until 480 days (Figures 4a–4d),
indicating that the migration rate of the CO2 plume was faster than the ambient groundwater
flow (1.94 cm/day). This is because the pumping activity developed an additional head
gradient, which was greater than ambient groundwater flow. After 480 days, the mass center
slightly moved back until 600 days (Figure 4e), implying that all movable CO2, which was
weighted CO2 mass at the plume front, was pumped out. After 600 days, continuous CO2
dissolution occurred at the plume rear, and the pumping activity at the front induced the
movement of the mass center slowly.
33
7.1.2 Induced Geochemical Reactions
The dissolution of gaseous CO2 into ambient groundwater increased HCO3-
concentration from 0.2 to 1 mol/L while decreasing the pH from 8.0 to 5.5 within the CO2
plume (Figures 5a and 5d). Concurrently, both carbonate and silicate minerals were either
dissolved or precipitated. In particular, dramatic changes in minerals were localized at both
rear and front margins of the CO2 plume where the vigorous advective mixing of
chemically different types of groundwater occurred.
Dissolutions in Calcite (CaCO3) and Magnesite (MgCO3) were distinct (Figures 5b
and 5c). Such dissolution released HCO3- into the ambient groundwater and subsequently
induced a positive feedback to lower pH. Calculation of the saturation index (SI) using the
initial concentration of ambient groundwater indicated that Calcite (SIcalcite=-3.1) was more
saturated than Magnesite (SImagnesite=-12.4); the initial concentrations of Ca2+ and Mg2+
were 3.310-3 mol/L and 4.210-12 mol/L, respectively, in the ambient groundwater. Due to
this reason, when CO2 was leaked, dissolution of Magnesite (-6 mol/m3) was greater than
that of Calcite (-4 mol/m3) within the CO2 plume. However, even if overall Magnesite
dissolution were greater than that of Calcite, the greatest change in mineral dissolution
appeared to Calcite (-8 mol/m3), focusing at the rear margin of the CO2 plume (Figure 5b).
This implies that the localized dissolution of Calcite was primarily induced by the mixing
of two chemically different types of groundwater, such as the ambient and CO2-dissolved
groundwater. Initially, the Ca2+concentration in the ambient groundwater was 3.310-3
mol/L (Figure 5e). Inside the CO2 plume, Calcite was dissolved and increased Ca2+
34
Fig
ure
5. D
istr
ibu
tio
n o
f pH
, ca
rbonat
e m
iner
als
(Cal
cite
and M
agnes
ite)
, an
d a
ssoci
ated
dis
solv
ed s
pec
ies
(HC
O3- ,
Ca2
+,
and
Mg
2+)
afte
r 3
65
day
s.
35
concentration 10-fold to 2.510-2 mol/L. At the rear margin of the CO2 plume, Ca2+
concentration was increased even more (to 3.110-2 mol/L). Overall, the distribution of
Ca2+ was similar to that of Calcite (Figures 5b and 5e), and distributions of both Mg2+ and
HCO3- were similar to that of Magnesite (Figures 5c, 5d, and 5f).
Patterns of dissolution and precipitation in silicate minerals were more complex than
those of carbonate minerals (Figures 6a–6c). CO2 leakage primarily induced the dissolution
of K-feldspar (KAlSi3O8) (Figure 6a); the greatest dissolution (3.510-2 mol/m3) occurred
at the rear of the CO2 plume, and the degree of dissolution gradually decreased as the plume
approached the municipal well. Dissolution of K-feldspar increased concentrations of K+,
SiO2(aq), and AlO2- in the groundwater (Figures 6d–6f). Nevertheless, distributions of such
species did not imitate the dissolution pattern of K-feldspar. Rather, SiO2(aq) and K+
showed the highest concentrations at the rear and front margins of the CO2 plume with
values of 1.810-4 mol/L and 3.510-4 mol/L, respectively (Figures 6d and 6e), but the
decrease in AlO2- concentration occurred uniformly throughout the CO2 plume (Figure 6f).
The discrepancy in patterns between K-feldspar and other dissolved species was
presumably caused by a combination of both dissolution and precipitation among various
silicate minerals such as Illite, Kaolinite, and Chlorite as described below.
For the distribution of Illite (K0.6Mg0.25Al1.8(Al0.5Si3.5O10)(OH)2), a small amount
(3.310-2 mol/m3) was precipitated throughout the CO2 plume (Figure 6b). However, at the
rear margin and immediately adjacent to the municipal well, a relatively large degree of
36
Fig
ure
6.
Dis
trib
uti
on
of
sili
cate
min
eral
s (K
-fel
dsp
ar,
Illi
te,
Chlo
rite
, an
d K
aoli
nit
e) a
nd
ass
oci
ated
dis
solv
ed s
pec
ies
(SiO
2(a
q),
K+, an
d A
lO2)
afte
r 365 d
ays.
37
Illite dissolution was predicted (-9.110-3 and -7.410-2 mol/m3, respectively). In contrast
to Illite, a small amount (-2.910-2 mol/m3) of Kaolinite was dissolved within the plume,
and reversely, a small amount of precipitation was predicted at the rear and front margins
(1.910-2 and 4.810-2 mol/m3, respectively) (Figure 6c). Presumably, the dissolution and
precipitation of Illite and Kaolinite would influence the distribution of SiO2(aq) and K+ in
addition to K-feldspar dissolution (Figures 6d and 6e). Finally, the behavior of individual
silicate mineral influenced the distribution of AlO2-, the concentration of which within the
plume was lower than that outside (Figure 6f). Overall, CO2 leakage induced dissolution or
precipitation of both carbonate and silicate minerals, and ultimately, changed both the
porosity and permeability of the shallow aquifer. The dissolution of carbonate minerals
primarily caused increase in permeability; permeability increased to 0.43% ( k′(%) =
kchanged−𝑘𝑖𝑛𝑖𝑡𝑖𝑎𝑙
𝑘𝑖𝑛𝑖𝑡𝑖𝑎𝑙×100 ) within the CO2 plume, and the most drastic increase (0.77%)
occurred at the rear margin of the CO2 plume (Figure 7).
Dissolution of Arsenopyrite, which was the primary reaction for predicting
carcinogenic health risk, occurred only within the CO2 plume with a dissolved amount of
4.8410-5 mol/m3 (Figure 8a). Similar to that of carbonate minerals, the greatest amount of
Arsenopyrite dissolution (8.5710-5 mol/m3) occurred at the rear margin of the plume due
to the vigorous mixing of two chemically different types of groundwater. Following Eq. 5,
oxidative dissolution of Arsenopyrite consumed 0.75 mole of O2(aq) and 1 mole of H+
while increasing the concentrations of total Arsenic (∑As), Fe2+, and HS- (Figures 8b–8e).
38
Figure 7. Distribution of altered permeability (%) after 365days. The changes in
permeability was predicted from the following equation (𝐤′ % =𝐤𝐜𝐡𝐚𝐧𝐠𝐞𝐝−𝒌𝒊𝒏𝒊𝒕𝒊𝒂𝒍
𝒌𝒊𝒏𝒊𝒕𝒊𝒂𝒍×
𝟏𝟎𝟎) because its variation was too small. Here, kinitial and kchanged indicates the initial
permeability and altered permeability at designated time, respectively
39
Figure 8. Distribution of Arsenopyrite dissolution and associated dissolved species
(∑As, Fe2+, HS-, O2(aq)) after 365 days.
40
In this study, ∑As represents the summation of primary species such as arsenite
(H3AsO30(aq)), which is the by-product of Arsenopyrite dissolution, as well as other arsenic
species such as H2AsO3-, H3AsO4(aq), H2AsO4
-, H2AsO42-, HAsO4
2-, HAsO2(aq), and
HAsS2(aq). Distribution of ∑As concentration mimics that of Arsenopyrite (Figures 8a and
8b); generally, ∑As concentration within the CO2 plume was greater than that outside,
while the rear margin revealed the highest concentration. However, the other associated
species such as Fe2+, HS-, and O2(aq) revealed relatively uniform distribution (Figures 8c–
8e). The difference between Arsenopyrite produced-species such as ∑As and other
associated species (e.g., Fe2+, HS-, and O2(aq)) presumably occurred due to the sorption
effect, which was accounted for in the linear Kd approach. The Kd was designated for
arsenic species only, and thus, as shown in Figure 8b, enrichment of ∑As concentration
occurred at the rear of the CO2 plume. In summary, the greatest concentration of ∑As was
4.910-7 mol/L at the rear margin of the plume and the average concentration of ∑As within
the plume was 2.9 mol/L (Figure 8b). The average concentrations of other species such as
Fe2+ and HS- were 3.910-7 mol/L and 3.610-7 mol/L, respectively (Figures 8c and 8d).
41
7.1.3 Health Risk Assessment of Carcinogenic Effect
To account for carcinogenic health risk, selected species such as pH, ∑As, and arsenite
(H3AsO30(aq)) concentrations were monitored at the municipal well for 100 years (Figure
9a). Dramatic changes in gaseous CO2 saturation (Sg), pH, and mass fraction of CO2
dissolved in groundwater (XCO2) predicted during 10 years were magnified at the small
window (Figure 9a). Depending on the profiles of both pH and dissolved arsenic species,
two stages (Stages I and II) were characterized. During Stage I (0–6.3 years), the leaked
CO2 plume, which existed as either gaseous CO2 (Sg, black dotted line) or dissolved CO2
(XCO2, purple dotted line), arrived at the municipal well approximately after 360 days. Once
the CO2 plume arrived at the well, immediate reduction of pH from 8.8 to 5.2 was observed,
while ∑As and H3AsO30(aq) concentrations sharply increased to 2.9310-7 and 1.3210-7
mol/L, respectively, exceeding the maximum contaminant level (MCL=1.3310-7 mol/L,
red dotted line) (EPA, 2008). After 2.1 years, Sg decreased to 0 at the municipal well,
implying that all movable gaseous CO2 was pumped out. Even after all gaseous CO2 had
been diminished due to the pumping activity, residually trapped CO2 remained within the
pores while dissolving into the groundwater. Due to the dissolution of residually trapped
CO2, the mass fraction of dissolved CO2 (XCO2) was invariant at 0.035 until 5 years; in this
shallow aquifer, flow caused by both ambient fresh groundwater and pumping activity
accelerated CO2 dissolution. Therefore, complete dissolution of residually trapped CO2
appeared at 6.3 years when XCO2 became 0 and pH returned to 8.2. In addition,
concentrations of ∑As and H3AsO30(aq) reached 3.7710-7 and 1.7010-7 mol/L,
42
Figure 9. (a) Changes in pH, total arsenic (∑As), and arsenite (H3AsO30) concentrations
monitored at the municipal well. The Maximum Contaminant Level (MCL) of arsenic
species is 1.33⨯10-7 mol/L. Gaseous CO2 saturation (Sg) and pH for 10 years was also
plotted at the magnified window. Two stages were characterized based on the behavior
of CO2 plume. (b) Frequency of predicted health risk accounting for both ∑As and
H3AsO30, and cumulative density functions representing probabilistic health risk. The
bin size of both histograms was designated to be 10-4, and thus, the number of bins was
10 and 20, respectively. Finally, the threshold of risk or the risk level of concern was
10-4 (red dotted line).
43
44
respectively. Immediately beginning with Stage II (6.3–100 years), elevated pH (8.2) of
groundwater inhibited the dissolution of Arsenopyrite, the reaction of which requires the
consumption of H+ (Eq. 5). Due to decreased Arsenopyrite dissolution, concentrations of
both ∑As and H3AsO30(aq) were stabilized and gradually decreased due to sorption on
aquifer media. The pH continuously decreased until 90 years. However, concentrations of
∑As and H3AsO30(aq) reached the background level (6.6010-8 and 3.0110-8 mol/L,
respectively) at 62 years.
Based on the simulated profiles of both ∑As and H3AsO30(aq), 𝐶̅ , which is the
maximum average concentration calculated from Eq. 8, was predicted, and the
carcinogenic health risk was quantified following the method described in Section 5.3.
Figure 9b shows histograms representing calculated frequencies of carcinogenic risk for
both ∑As (blue bar) and H3AsO30(aq) (green bar). The carcinogenic risk predicted from
H3AsO30(aq) profile, which revealed relatively low concentrations, showed a mean, median,
and standard deviation of 4.0010-4, 4.3010-4, and 1.6910-4, respectively. For risk
predicted from ∑As concentration profile, the mean, median and standard deviation was
8.9410-4, 9.6010-4, and 3.7710-4, respectively. Cumulative density functions (blue and
green lines) were also plotted together with the risk level of concern (10-4) (EPA, 2001;
McSlarrow, 1991). From the cumulative density functions, the risk, which exceeds the risk
level of concern, can be considered to have carcinogenic potential after chronic exposure
to arsenic-contaminated groundwater. As shown. the risk level predicted from both ∑As
and H3AsO30(aq) exceeded the risk level (red dotted line) of concern.
45
7.2 Sensitivity Studies
7.2.1 Effect of Sorption Intensity (Kd) (Cases 2–4)
In this sensitivity study, the intensity of the sorption effect (Kd of 25, 50, and 100 L/kg)
on ∑As was evaluated while gaseous CO2 was leaked into the shallow aquifer (Table 6).
As shown in Figure 4, once gaseous CO2 was leaked from unidentifiable pathways, it
migrated with the ambient groundwater. Here, characteristics of gaseous CO2 plume such
as its size, shape, migrating velocity, and gas saturation (Sg) were influenced by multiphase
parameters such as capillary pressure and relative permeability (Table 1); the role of Kd on
gaseous CO2 transport was minimal because the sorption typically accounted for the
movement of dissolved species within geologic media. Due to this reason, regardless of the
variation in Kd, the distribution and behavior of gaseous CO2 plume was not affected, and
therefore, the CO2 plume remained essentially the same in all cases as shown in Figure 4.
However, within the gaseous CO2 plume various geochemical processes occurred,
including reductions in pH, dissolution of Arsenopyrite, and interactions between dissolved
species. In particular, variation in Kd was anticipated to affect the behavior of dissolved
arsenic species after all gaseous CO2 was pumped out or dissolved into the groundwater.
In Stage I, the gaseous CO2 plume arrived at the municipal well after 350 days (0.95
year) for all cases (Figure 10a). The elevated concentration (2.910-7 mol/L) of ∑As
occurred simultaneously in all cases because the source of arsenic species was the
46
Figure 10. (a) ∑As profiles with different Kd from 25, 50, to 100 L/kg-1 (Cases 2-4) and
(b) probabilistic health risk plotted as cumulative density functions. Red dotted-line
represents the risk level of concern (10-4)
47
48
dissolution of Arsenopyrite, which resulted from CO2 dissolution. Despite the variation in
Kd, the arrival times of ∑As at the municipal well were the same because the sorption did
not affect the migration of the gaseous CO2 plume. Therefore, the ∑As profiles evolved
similarly until approximately 30 years (the middle of Stage II) when its concentration
reached a maximum (3.910-7 mol/L). After the arrival of the ∑As peak, the differences
between ∑As profiles were amplified until the ∑As concentration reached the background
level (6.710-8 mol/L). In detail, differences in the slopes of ∑As profiles were small
immediately after the arrival of the ∑As peak (30–40 years), but the discrepancy was
amplified from 40 years while the slopes for ∑As profiles sharply dropped. Differences in
∑As profiles were attributed to the degree of Kd, which determined the amount of arsenic
adsorbed to aquifer media, especially at the rear margin of the CO2 plume; at this location,
the highest ∑As concentration occurred due to the mixing of two chemically different types
of groundwater as shown in Figure 8b.
In these simulations, with increasing Kd, more arsenic was adsorbed to the aquifer
media, and the migration of ∑As was therefore retarded. In other words, stronger
retardation caused ∑As concentration to be maintained higher and longer in the aquifer,
and therefore, the arrival of ∑As concentration at the background level was delayed. For
example, in Case 2 (Kd=25 L/kg), ∑As concentration revealed the earliest recovery (69.3
years) at the municipal well (Figure 10a). As Kd increased to 50 and 100 L/kg, the recovery
time was delayed to 72.1 and 73.9 years, respectively.
49
Figure 10b represents the predicted probabilistic health risk for Cases 2–4. While
calculating the health risk, the C̅, the peak calculated from the moving average of ∑As
concentration by adopting the designated interval of ED (30 years), influenced the health
risk most significantly based on Eqs. 8, 9, and 10. Since ∑As profiles revealed similar
patterns with the same peak values while the only difference being the recovery time, the
calculated C̅ values for Cases 2–4 were almost the same (3.910-7 mol/L). Consequently,
predicted carcinogenic health risk for humans was almost identical to variation in Kd. This
result implies that variation in Kd was a less influential parameter for assessing health risk
for arsenic species. This was because variation in Kd did not affect the characteristics of the
CO2 plume such as size, shape, and migration velocity, which determined the dissolved
amount or reaction rate for arsenic sources such as Arsenopyrite. Due to this reason,
additional simulations were conducted and described in the following section after varying
parameters (CO2 leakage rate and aquifer permeability), which directly affected the size of
the CO2 plume and its migration rate.
50
7.2.2 Effect of CO2 Leakage Rate (QL) (Cases 5–7)
In Cases 5–7, the effect of CO2 leakage rate (QL) on the quantification of carcinogenic
health risk was evaluated. Figures 11a–11c show the distribution of the gaseous CO2 plume
after CO2 leakage was stopped at 1 year, and Figures 11d–11f present the distribution of
∑As concentration after 20 years. As the QL increased from 0.020, 0.025, to 0.030 kg/s, the
CO2 plume approached the municipal well more closely; the calculated mass center of the
CO2 plume, plotted as black circles at a 120-day interval, moved rapidly as the QL increased.
In addition, adjacent to the CO2 leakage point, CO2 saturation was elevated from 0.27 to
0.30, implying occurrence of active dissolution of gaseous CO2. As described before,
although the moveable gaseous CO2 plume was pumped out after approximately 2 years,
residually trapped CO2 still remained in the pores, concurrently dissolving into the ambient
groundwater until 8 years. Increases in dissolved CO2 concentration due to dissolution of
residually trapped CO2 and resulting low-pH groundwater influenced the distribution of
∑As concentration even after all gaseous CO2 was pumped out (Figures 11d–11f). For
example, the intensity of the QL governed the size of the CO2 plume where the active
dissolution of Arsenopyrite, the source of ∑As, primarily occurred. Therefore, as QL
increased from 0.020, 0.025, to 0.030 kg/s, the size of ∑As plume expanded at 20 years.
Additionally, the effects of dispersion, diffusion, and sorption were amplified while the
∑As plume migrated in the ambient groundwater.
The ∑As profiles monitored at the municipal well revealed the drastic increase in
∑As concentration immediately after the arrival of the ∑As plume at 1.31, 1.07, and 0.98
51
Fig
ure
11. (a
), (
b),
an
d (
c) S
atura
tion o
f le
aked
CO
2 a
t dif
fere
nt
leak
age
rate
aft
er 1
yea
r (3
65
day
s) (
Cas
es 5
-7).
(d
), (
e),
and
(f)
Dis
trib
uti
on
of
∑A
s co
nce
ntr
atio
n i
nduce
d b
y l
eaked
CO
2 a
t dif
fere
nt
leak
age
inte
nsi
ty (
QL)
afte
r 2
0 y
ears
(C
ase
5-7
).
52
years for Cases 5, 6, and 7, respectively (Figure 12a). The arrival time of the ∑As plume
coincided with that of the gaseous CO2 plume, implying that multi-phase migration of CO2
governed the movement of dissolved ∑As. The size of the ∑As peaks, all of which were
over the MCL, increased with QL (1.5610-7, 2.7510-7, and 4.0710-7 mol/L for Cases 5,
6, and 7, respectively). In addition, its peak was maintained for a longer time with increased
QL (13.5, 16.6, and 23.0 years for Cases 5, 6, and 7, respectively). Figure 12b shows the
calculated probabilistic health risk. While calculating the health risk using Eqs. 9 and 10,
an important parameter was 𝐶̅, which was the maximum average of arsenic concentration
monitored at the municipal well (Eq. 8). The size of QL varied 𝐶̅ as shown in Figure 12a,
which directly affected the risk prediction. Therefore, as the ∑As concentration increased,
the health risk for humans increased accordingly. Especially, the medians of Cases 5, 6, and
7, located at the half percentile in variability, were 3.3910-4, 6.5310-4, and 9.5610-4,
respectively, all of which exceeded the risk level of concern (10-4).
53
Figure 12. (a) ∑As profiles with different leakage intensity (QL) from 0.020, 0.025, to
0.030 kg/s (Cases 5-7) and (b) probabilistic health risk for Case 5-7 plotted as
cumulative density functions (CDF). Red dotted-line represents the risk level of concern
(10-4).
54
55
7.2.3 Effect of Horizontal Permeability (kh) (Cases 8–10)
From two previous sensitivity studies, it was revealed that the driving force of the ∑As
plume was essentially the movement of the gaseous CO2 plume; depending on the size of
the CO2 plume, the amount of dissolved ∑As was determined. In this study, the magnitude
of horizontal permeability (kh), which altered the velocity of ambient groundwater flow,
varied from 0.210-13, 1.010-13, to 5.010-13 m2 (Cases 8–10, respectively). As kh increased,
the ambient groundwater rate increased from 0.39, 1.94 to 9.71 cm/day, which induced a
change in the shape of the CO2 plume from oval to flat (Figures 13a–13c).
Typically, the shape and movement of the gaseous CO2 plume within the groundwater
aquifer is governed by the balance between buoyancy and viscous forces due to the density
contrast of these two fluids. Even, the density contrast is amplified as CO2 leaks into the
shallower aquifer, because CO2 density radically decreases while approaching the surface.
The magnitude of the buoyancy number (Nb =𝑘𝑣𝐿∆𝜌𝑔
Hvμ) reflects the change in CO2 plume
shape, where, g is gravitational acceleration, L and H are the length and height of the model
domain, respectively, v is the ambient groundwater flow rate, and 𝜇 is dynamic viscosity
of CO2 (1.4710-5 Pa∙s). Finally, ∆𝜌 is the density difference between groundwater (1,000
kg/m3) and CO2 (1.842 kg/m3) (Span and Wagner, 1996). The calculated Nb of Cases 8, 9,
and 10 was 740.30, 148.06, and 29.61, respectively, implying that the buoyancy force
acting on CO2 plume increased as kh decreased.
56
Fig
ure
13.
(a),
(b
), a
nd (
c) S
atura
tion o
f le
aked
CO
2 a
t dif
fere
nt
hori
zonta
l per
mea
bil
ity (
kh)
afte
r 1
yea
r (3
65
day
s)
(Cas
es 8
-10
). (d
), (e
), an
d (f
) D
istr
ibuti
on of
∑A
s C
once
ntr
atio
n in
duce
d by le
aked
C
O2 at
dif
fere
nt
ho
rizo
nta
l
per
mea
bil
ity (
kh)
afte
r 20 y
ears
(C
ases
8-1
0).
57
Even if the velocity of the CO2 plume was the lowest in Case 8 (or the largest Nb), the
CO2 plume swept the largest area, covering the entire thickness of the aquifer; the
calculated mass center moved to the middle height of the aquifer (Figure 13a). Due to large
coverage of the CO2 plume, the size of the corresponding ∑As plume was the largest in
Case 8 (Figure 13d). In contrast, Case 10 with high kh accelerated the advective
groundwater flow and dispersion, resulting in the flattened shape of the CO2 plume sinking
to the bottom of the aquifer (Figure 13c). Due to the poor sweeping efficiency of the CO2
plume, the ∑As plume only migrated beneath the municipal well.
The ∑As concentrations at the municipal well for Cases 8–10 are plotted in Figure
14a. The arrival time of ∑As concentration was the latest (4.95 years) in Case 8 due to a
low kh. However, due to large coverage of the ∑As plume, the ∑As concentration at the
municipal well continuously increased to 3.7710-7 mol/L until 190 years. In contrast, for
Case 10, the arrival time of the ∑As plume was the shortest (0.95 years), reaching a peak
of 1.7310-7 mol/L. However, due to dominance of high advective flow, the ∑As plume
was flattened below the municipal well. Therefore, the amount of ∑As captured from the
municipal well decreased soon, reaching the background level (0.6710-7 mol/L) only after
25 years. These results imply that the location of the wellbore (e.g., fully or partially
penetrating well, the location of screen interval) and the size of the capture zone (e.g., the
pumping capacity) are important characteristics for governing ∑As concentration at the
well.
58
Figure 14. (a) ∑As profiles with different horizontal permeability (kh) from 0.2×10-13,
1.0×10-13, to 5.0×10-13 m2 (Cases 8-10) and (b) probabilistic health risk for Case 8-10
plotted as cumulative density functions (CDF). Red dotted-line represents the risk level
of concern (10-4).
59
60
The calculated health risk for each case is plotted at Figure 14b. As expected, Case 10,
which showed the smallest breakthrough of the ∑As concentration (e.g., the smallest 𝐶̅),
revealed the lowest risk. Both Case 8 and 9 showed almost equivalent high-risk prediction
even if the profile of the ∑As concentration appeared differently at the municipal well
(Case 8: ~80 years and Case 9: ~180 years). While accounting for the risk assessment, the
exposure duration (ED) was chosen to be 30 years in this work (EPA, 1989). Following Eq.
8, the 𝐶̅ calculated from the ∑As concentration profile is typically dependent on the
duration of ED (Maxwell et al., 2008; Siirila and Maxwell, 2012; Siirila et al., 2012). For
example, the 𝐶̅ can decrease when the break-through of ∑As concentration is shorter than
the ED. However, when the breakthrough of the ∑As concentration is sufficiently longer
than the chosen ED, the 𝐶̅ does not change. Similarly, in both Case 8 and 9, the profiles
of ∑As concentration at the municipal well were sufficiently longer than the chosen ED
(Figure 14a). Due to this reason, 𝐶̅ for these two cases were similar each other, and
subsequently, the predicted risk levels only showed the slight difference.
61
8 Conclusion
Even if there are no direct evidences showing that the stored CO2 has leaked to the
shallow aquifers from any major CO2 injection demonstration sites (Finley et al., 2013;
Mathieson et al., 2011; Torp and Gale, 2004), there exist a few natural analog sites
indicating that naturally stored CO2 has leaked through pre-existing fault systems (Farrar
et al., 1995; Han et al., 2013; Keating et al., 2013b; Kling et al., 1987). These natural sites
where CO2 leakage are primarily driven by geothermal or tectonic activities are typically
less populated with lack of concern in potable groundwater resources. However, as the
number of CO2 injection demonstration activities is growing, the secondary contamination
of leaked CO2 to the shallow potable aquifer becomes an important issue. In this study, with
presence of Arsenic-bearing minerals in the aquifer, it is suggested that water-rock
interactions induced by CO2 leakage could mobilize arsenic species to the shallow potable
aquifer. Therefore, without proper treatments, any residences who continuously utilize
these groundwater resources could have great probability to expose cancer-related diseases.
Throughout this study, we developed a 2D confined aquifer model where gaseous CO2
was leaked, and a nearby municipal well concurrently pumped out both leaked CO2 and
groundwater. Immediately after a leaked CO2 plume arrived at the municipal well,
concentrations of As species increased together, indicating occurrence of active
Arsenopyrite dissolution in the aquifer media. Subsequently, by analyzing As concentration
62
from the municipal well, the carcinogenic health risk was quantified. The processes
characterizing the movement of gaseous CO2 and associated CO2-water-rock reactions
were simulated with the multi-phase, multi-species reactive transport model, and
subsequent carcinogenic health risks were predicted with probabilistic approach. The
simulation results revealed that the movement of leaked CO2 plume was governed by local
flow fields within the shallow potable aquifer; three driving forces such as ambient
groundwater flow, CO2 leakage-driven flow, and pumping-driven flow were characterized.
This complex flow field governed chemical reactions, resulting that the most drastic
increase (0.77%) in permeability occurred at the rear margin of the CO2 plume where the
vigorous mixing between the ambient groundwater and CO2-dissolved fluid accelerated
dissolution of the minerals. Additionally, sensitivity studies were conducted while varying
the sorption intensity, leakage rate of CO2, and horizontal permeability.
However, there are still some limitations within the numerical simulation and risk
assessment. First, as a part of numerical simulation, we approximated the sorption of
arsenic via the Kd approach. But, to represent the sorption of clay minerals or iron-oxides
with greater accuracy, modelling must be conducted by combined reactive transport and
surface complex model. Secondly, for the risk assessment, several risk parameters are based
on the EPA guidelines for a typical US citizen. The difference in life styles and average
physique are different from country to country. Therefore, for a risk assessment in countries
aside from the US, new standard parameters should be derived based on its citizens, and
used. Lastly, the risk for the arsenic contamination was quantified using the total arsenic
63
concentration, considering arsenite (As(III)), arsenate (As(V)) and the other species. The
toxicity of each arsenic species is different from each other, but the risk quantification was
conducted with only cancer potency factor (CPF) for total arsenic. For the accurate
quantification of risk for each species, consideration of the CPF for each species is required.
In summary, key factors that exacerbate the secondary contamination of arsenic
species at the municipal well, were physical characteristics of CO2 plume such as shape,
size, and migration velocity of CO2 plume; these physical characteristics govern the area
where Arsenopyrite dissolves, subsequently affecting As concentration. Furthermore, the
size of capture zone (e.g., screen interval, pump capacity) also governed the As
concentration in the municipal well. Therefore, when the secondary contamination occurs
within the shallow potable aquifer, it is suggested that the aquifer characteristics as well as
the amount of leaked CO2 and its plume size should be evaluated to develop a proper
remediation protocol. At the same time, the prompt shut-down of any nearby municipal
wells could minimize any potential hazards.
64
9 References
André, L., Azaroual, M., Bernstone, C., Wittek, A., 2014. Modeling the geochemical
impact of an injection of CO2 and associated reactive impurities SO2 and O2 into
a saline reservoir. Transport in Porous Media 108, 185-205.
Apps, J.A., Zheng, L., Spycher, N., Birkholzer, J.T., Kharaka, Y., Thordsen, J., Kakouros,
E., Trautz, R., 2011. Transient changes in shallow groundwater chemistry during
the MSU ZERT CO2 injection experiment. Energy Procedia 4, 3231-3238.
ATSDR, 2015 Support document to the 2015 priority list of hazardous substances that will
be candidates for toxicological profiles, Agency for Toxic Substances and Disease
Registry, Atlanta, USA.
Bachu, S., 2008. Legal and regulatory challenges in the implementation of CO2 geological
storage: An Alberta and Canadian perspective. Int J Greenh Gas Con 2, 259-273.
Berman, E., Berman, E., 1980 Toxic metals and their analysis.
Bethke, C.M., Brady, P.V., 2000. How the Kd approach undermines ground water cleanup.
Groundwater 38, 435-443.
Bhattacharya, P., Chatterjee, D., Jacks, G., 1997. Occurrence of arsenic-
contaminatedgroundwater in alluvial aquifers from delta plains, eastern india:
Options for safe drinking water supply. International Journal of Water Resources
Development 13, 79-92.
65
Birkholzer, J.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
Greenh Gas Con 3, 181-194.
Bose, P., Sharma, A., 2002. Role of iron in controlling speciation and mobilization of
arsenic in subsurface environment. Water Res 36, 4916-4926.
Carroll, S.A., Keating, E., Mansoor, K., Dai, Z.X., Sun, Y.W., Trainor-Guitton, W., Brown,
C., Bacon, D., 2014. Key factors for determining groundwater impacts due to
leakage from geologic carbon sequestration reservoirs. Int J Greenh Gas Con 29,
153-168.
Carruthers, C.I.A. (2016) Metal mobility in sandstones and the potential environmental
impacts of offshore geological CO2 storage, The University of Edinburgh.
Chen, M., Mao, J., Li, C., Zhang, Z., Dang, Y., 2015. Re–Os isochron ages for arsenopyrite
from Carlin-like gold deposits in the Yunnan–Guizhou–Guangxi “golden triangle”,
southwestern China. Ore Geology Reviews 64, 316-327.
Chowdhury, U.K., Biswas, B.K., Chowdhury, T.R., Samanta, G., Mandal, B.K., Basu, G.C.,
Chanda, C.R., Lodh, D., Saha, K.C., Mukherjee, S.K., 2000. Groundwater arsenic
contamination in Bangladesh and West Bengal, India. Environmental health
perspectives 108, 393.
Cutter, G.A., 1992. Kinetic controls on metalloid speciation in seawater. Marine Chemistry
40, 65-80.
66
Davis, J., Coston, J., Kent, D., Fuller, C., 1998. Application of the surface complexation
concept to complex mineral assemblages. Environ Sci Technol 32, 2820-2828.
Dockrill, B., Shipton, Z.K., 2010. Structural controls on leakage from a natural CO2
geologic storage site: Central Utah, USA. Journal of Structural Geology 32, 1768-
1782.
Doughty, C., Freifeld, B.M., Trautz, R.C., 2008. Site characterization for CO2 geologic
storage and vice versa: the Frio brine pilot, Texas, USA as a case study. Environ
Geol 54, 1635-1656.
Duan, Z., Sun, R., 2003. An improved model calculating CO2 solubility in pure water and
aqueous NaCl solutions from 273 to 533 K and from 0 to 2000 bar. Chemical
geology 193, 257-271.
EPA, U.S., 1989 Risk Assessment Guidance for Superfund. Response, O.o.E.a.R. (ed), U.S.
Environmental Protection Agency, Washington D.C.
EPA, U.S., 2001 Risk Assessment Guildance For Superfund. Response, O.o.E.a.R. (ed),
U.S Environmental Protection Agency, Washington D.C.
EPA, U.S., 2004 Risk Assessment Guidance for Superfund. Response, O.o.E.a.R. (ed), U.S.
Environmental Protection Agency, Washington D.C.
EPA, U.S., 2005 Guidelines for Carcinogen Risk Assessment. Agency, U.S.E.P. (ed), Risk
Assessment Forum, Washington, DC.
67
EPA, U.S., 2008 National Primary Drinking Water Regulations, United States
Environmental Protection Agency.
Fahrner, S., Schafer, D., Dethlefsen, F., Dahmke, A., 2012. Reactive modelling of CO2
intrusion into freshwater aquifers: current requirements, approaches and limitations
to account for temperature and pressure effects. Environ Earth Sci 67, 2269-2283.
Farrar, C., Sorey, M., Evans, W.C., Howle, J., Kerr, B., Kennedy, B.M., King, C.-Y.,
Southon, J., 1995. Forest-killing diffuse CO2 emission at Mammoth Mountain as a
sign of magmatic unrest. Nature 376, 675-678.
Finley, R.J., Frailey, S.M., Leetaru, H.E., Senel, O., Couëslan, M.L., Scott, M., 2013. Early
operational experience at a one-million tonne CCS demonstration project, Decatur,
Illinois, USA. Energy Procedia 37, 6149-6155.
Guyant, E., Han, W.S., Kim, K.-Y., Park, E., Yun, S.-T., 2016. Leakage and pressurization
risk assessment of CO2 reservoirs: A metamodeling modeling approach. Int J
Greenh Gas Con 54, 345-361.
Han, W.S., Lu, M., Mcpherson, B.J., Keating, E., Moore, J., Park, E., Watson, Z., Jung,
N.H., 2013. Characteristics of CO2‐driven cold‐water geyser, Crystal Geyser in
Utah: experimental observation and mechanism analyses. Geofluids 13, 283-297.
Han, W.S., Watson, Z., Kampman, N., Grundl, T., Graham, J.P., Keating, E.H., 2017.
Periodic changes in effluent chemistry at cold-water geyser: Crystal geyser in Utah.
Journal of Hydrology 550, 54-64.
68
Hebbar Annapoorna, J.M., 2016. Arsenic contamination in groundwater of the areas
surrounding Ingaldhal, Chitradurga district, Karnataka state. International Journal
of Geology, Earth & Environmental Science 6, 1-7.
Helgeson, H.C., Kirkham, D.H., Flowers, G.C., 1981. Theoretical prediction of the
thermodynamic behavior of aqueous electrolytes by high pressures and
temperatures; IV, Calculation of activity coefficients, osmotic coefficients, and
apparent molal and standard and relative partial molal properties to 600 degrees C
and 5kb. American journal of science 281, 1249-1516.
Hills, E.S., Thomas, D., 1945. Fissuring in sandstones. Economic Geology 40, 51-61.
Hu, Q., Sun, G., Gao, X., Zhu, Y., 2012. Conversion, sorption, and transport of arsenic
species in geological media. Appl Geochem 27, 2197-2203.
IRIS, I.R.I.S., 1991 Chemical Assessment Summary, Arsenic, inorganic; CASRN 7440-38-
2. Assessment, N.C.f.E. (ed), U.S. Environmental Protection Agency.
Jain, C.K., Ali, I., 2000. Arsenic: Occurrence, toxicity and speciation techniques. Water Res
34, 4304-4312.
Kampman, N., Bickle, M., Becker, J., Assayag, N., Chapman, H., 2009. Feldspar
dissolution kinetics and Gibbs free energy dependence in a CO2-enriched
groundwater system, Green River, Utah. Earth and Planetary Science Letters 284,
473-488.
69
Kampman, N., Bickle, M., Maskell, A., Chapman, H., Evans, J., Purser, G., Zhou, Z.,
Schaller, M., Gattacceca, J.C., Bertier, P., 2014. Drilling and sampling a natural
CO2 reservoir: implications for fluid flow and CO2-fluid–rock reactions during
CO2 migration through the overburden. Chemical Geology 369, 51-82.
Keating, E.H., Fessenden, J., Kanjorski, N., Koning, D.J., Pawar, R., 2009. The impact of
CO2 on shallow groundwater chemistry: observations at a natural analog site and
implications for carbon sequestration. Environ Earth Sci 60, 521-536.
Keating, E.H., Hakala, J.A., Viswanathan, H., Carey, J.W., Pawar, R., Guthrie, G.D.,
Fessenden-Rahn, J., 2013a. CO2 leakage impacts on shallow groundwater: Field-
scale reactive-transport simulations informed by observations at a natural analog
site. Appl Geochem 30, 136-147.
Keating, E.H., Newell, D.L., Viswanathan, H., Carey, J.W., Zyvoloski, G., Pawar, R., 2013b.
CO2/Brine transport into shallow aquifers along fault zones. Environ Sci Technol
47, 290-297.
Kharaka, Y.K., Cole, D.R., Hovorka, S.D., Gunter, W.D., Knauss, K.G., Freifeld, B.M.,
2006. Gas-water-rock interactions in Frio Formation following CO2 injection:
Implications for the storage of greenhouse gases in sedimentary basins. Geology 34,
577-580.
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., 2010. Changes in the
70
chemistry of shallow groundwater related to the 2008 injection of CO2 at the ZERT
field site, Bozeman, Montana. Environ Earth Sci 60, 273-284.
Kling, G.W., Clark, M.A., WAGNER, G.N., COMPTON, H.R., HUMPHREY, A.M.,
DEVINE, J.D., EVANS, W.C., Lockwood, J.P., Tuttle, M.L., KOENIGSBERG, E.J.,
1987. The 1986 lake nyos gas disaster in cameroon, west Africa. Science 236, 169-
175.
Knauss, K.G., Johnson, J.W., Steefel, C.I., 2005. Evaluation of the impact of CO2, co-
contaminant gas, aqueous fluid and reservoir rock interactions on the geologic
sequestration of CO2. Chemical Geology 217, 339-350.
Larsen, D., Gentry, R.W., Solomon, D.K., 2003. The geochemistry and mixing of leakage
in a semi-confined aquifer at a municipal well field, Memphis, Tennessee, USA.
Appl Geochem 18, 1043-1063.
Lu, P., Zhu, C., 2010. Arsenic Eh–pH diagrams at 25°C and 1 bar. Environ Earth Sci 62,
1673-1683.
Mathieson, A., Midgely, J., Wright, I., Saoula, N., Ringrose, P., 2011. In Salah CO2 Storage
JIP: CO2 sequestration monitoring and verification technologies applied at Krechba,
Algeria. Energy Procedia 4, 3596-3603.
Matschullat, J., 2000. Arsenic in the geosphere - a review. Sci Total Environ 249, 297-312.
Maxwell, R.M., Carle, S.F., Tompson, A.F., 2008. Contamination, risk, and heterogeneity:
71
on the effectiveness of aquifer remediation. Environ Geol 54, 1771-1786.
Maxwell, R.M., Pelmulder, S.D., Tompson, A.F.B., Kastenberg, W.E., 1998. On the
development of a new methodology for groundwater-driven health risk assessment.
Water Resources Research 34, 833-847.
McSlarrow, K.E., 1991 National Oil and Hazardous Substances Pollution Contingency
Plan, United States Environmental Protection Agency, JSTOR.
Means, B., 1989 Risk-Assessment Guidance for Superfund. vol 1. Human Health
Evaluation Manual. (part A Interim report (final)), Environmental Protection
Agency, Washington, DC (USA). Office of Solid Waste and Emergency Response.
Meharg, A.A., Rahman, M.M., 2003. Arsenic contamination of Bangladesh paddy field
soils: implications for rice contribution to arsenic consumption. Environ Sci
Technol 37, 229-234.
Metz, B., Davidson, O., de Coninck, H., Loos, M., Meyer, L., 2005 IPCC special report on
carbon dioxide capture and storage, Intergovernmental Panel on Climate Change.
Navarre-Sitchler, A.K., Maxwell, R.M., Siirila, E.R., Hammond, G.E., Lichtner, P.C., 2013.
Elucidating geochemical response of shallow heterogeneous aquifers to CO2
leakage using high-performance computing: Implications for monitoring of CO2
sequestration. Adv Water Resour 53, 45-55.
NETL, 2013 United States 2012 Carbon Utilization And Storage Atlas, National Energy
72
Technology Laboratory.
Nordstrom, D.K., 2002. Worldwide occurrences of arsenic in ground water. Science 296,
2143-2145.
Oldenburg, C.M., Doughty, C., Peters, C.A., Dobson, P.F., 2012. Simulations of long‐
column flow experiments related to geologic carbon sequestration: Effects of outer
wall boundary condition on upward flow and formation of liquid CO2. Greenhouse
Gases: Science and Technology 2, 279-303.
Palandri, J.L., Kharaka, Y.K., 2004 A compilation of rate parameters of water-mineral
interaction kinetics for application to geochemical modeling, DTIC Document.
Parry, W., Forster, C.B., Evans, J.P., Bowen, B.B., Chan, M.A., 2007. Geochemistry of CO2
sequestration in the Jurassic Navajo Sandstone, Colorado Plateau, Utah.
Environmental Geosciences 14, 91-109.
Pruess, K., 2008. Leakage of CO2 from geologic storage: Role of secondary accumulation
at shallow depth. Int J Greenh Gas Con 2, 37-46.
Pruess, K., Spycher, N., 2007. ECO2N–A fluid property module for the TOUGH2 code for
studies of CO2 storage in saline aquifers. Energ Convers Manage 48, 1761-1767.
Reichle, D., Houghton, J., Kane, B., Ekmann, J., 1999 Carbon sequestration research and
development, Oak Ridge National Lab., TN (US); National Energy Technology
Lab., Pittsburgh, PA (US); National Energy Technology Lab., Morgantown, WV
73
(US).
Renken, R.A., Cunningham, K.J., Zygnerski, M.R., Wacker, M.A., Shapiro, A.M., Harvey,
R.W., Metge, D.W., Osborn, C.L., Ryan, J.N., 2005. Assessing the vulnerability of
a municipal well field to contamination in a karst aquifer. Environmental &
Engineering Geoscience 11, 319-331.
Rodríguez-Lado, L., Sun, G., Berg, M., Zhang, Q., Xue, H., Zheng, Q., Johnson, C.A.,
2013. Groundwater arsenic contamination throughout China. Science 341, 866-868.
Sakata, M., 1987. Relationship between adsorption of arsenic (III) and boron by soil and
soil properties. Environ. Sci. Technol.;(United States) 21.
Shipton, Z.K., Evans, J.P., Kirschner, D., Kolesar, P.T., Williams, A.P., Heath, J., 2004.
Analysis of CO2 leakage through ‘low-permeability’ faults from natural reservoirs
in the Colorado Plateau, east-central Utah. Geological Society, London, Special
Publications 233, 43-58.
Siirila, E.R., Maxwell, R.M., 2012. A new perspective on human health risk assessment:
Development of a time dependent methodology and the effect of varying exposure
durations. Sci Total Environ 431, 221-232.
Siirila, E.R., Navarre-Sitchler, A.K., Maxwell, R.M., McCray, J.E., 2012. A quantitative
methodology to assess the risks to human health from CO2 leakage into
groundwater. Adv Water Resour 36, 146-164.
74
Smedley, P.L., Kinniburgh, D.G., 2002. A review of the source, behaviour and distribution
of arsenic in natural waters. Appl Geochem 17, 517-568.
Sonnenthal, E., Ito, A., Spycher, N., Yui, M., Apps, J., Sugita, Y., Conrad, M., Kawakami,
S., 2005. Approaches to modeling coupled thermal, hydrological, and chemical
processes in the drift scale heater test at Yucca Mountain. International Journal of
Rock Mechanics and Mining Sciences 42, 698-719.
Span, R., Wagner, W., 1996. A new equation of state for carbon dioxide covering the fluid
region from the triple-point temperature to 1100 K at pressures up to 800 MPa. J
Phys Chem Ref Data 25, 1509-1596.
Steefel, C.I., Lasaga, A.C., 1994. A coupled model for transport of multiple chemical
species and kinetic precipitation/dissolution reactions with application to reactive
flow in single phase hydrothermal systems. American Journal of science 294, 529-
592.
Strandli, C.W., Benson, S.M., 2013. Identifying diagnostics for reservoir structure and CO2
plume migration from multilevel pressure measurements. Water Resources
Research 49, 3462-3475.
Tester, J.W., Worley, W.G., Robinson, B.A., Grigsby, C.O., Feerer, J.L., 1994. Correlating
quartz dissolution kinetics in pure water from 25 to 625 C. Geochim Cosmochim
Ac 58, 2407-2420.
Thornburg, K., Sahai, N., 2004. Arsenic occurrence, mobility, and retardation in sandstone
75
and dolomite formations of the Fox River Valley, Eastern Wisconsin. Environ Sci
Technol 38, 5087-5094.
Torp, T.A., Gale, J., 2004. Demonstrating storage of CO2 in geological reservoirs: the
Sleipner and SACS projects. Energy 29, 1361-1369.
Trautz, R.C., Pugh, J.D., Varadharajan, C., Zheng, L.G., Bianchi, M., Nico, P.S., Spycher,
N.F., Newell, D.L., Esposito, R.A., Wu, Y.X., 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.
Valocchi, A.J., 1984. Describing the transport of ion‐exchanging contaminants using an
effective Kd approach. Water Resources Research 20, 499-503.
Van Genuchten, M.T., 1980. A closed-form equation for predicting the hydraulic
conductivity of unsaturated soils. Soil Sci Soc Am J 44, 892-898.
Viswanathan, H., Dai, Z., Lopano, C., Keating, E., Hakala, J.A., Scheckel, K.G., Zheng, L.,
Guthrie, G.D., Pawar, R., 2012. Developing a robust geochemical and reactive
transport model to evaluate possible sources of arsenic at the CO2 sequestration
natural analog site in Chimayo, New Mexico. Int J Greenh Gas Con 10, 199-214.
Wang, S., Jaffe, P.R., 2004. Dissolution of a mineral phase in potable aquifers due to CO2
releases from deep formations; effect of dissolution kinetics. Energ Convers
Manage 45, 2833-2848.
76
WHO, 2012. Chemical Agents and Related Occupations - A Review of Human Carcinogens.
International Agency for Research on Cancer, France.
Xiao, T., Dai, Z., Viswanathan, H., Hakala, A., Cather, M., Jia, W., Zhang, Y., McPherson,
B., 2017. Arsenic mobilization in shallow aquifers due to CO2 and brine intrusion
from storage reservoirs. Sci Rep-Uk 7, 2763.
Xu, T., 2008 TOUGHREACT User's Guide: A Simulation Program for Non-isothermal
Multiphase Reactive Geochemical Transport in Variably Saturated Geologic Media,
V1.2.1.
Xu, T., Kharaka, Y.K., Doughty, C., Freifeld, B.M., Daley, T.M., 2010. Reactive transport
modeling to study changes in water chemistry induced by CO2 injection at the Frio-
I Brine Pilot. Chemical Geology 271, 153-164.
Xu, T.F., Sonnenthal, E., Spycher, N., Pruess, K., 2006. TOUGHREACT - A simulation
program for non-isothermal multiphase reactive geochemical transport in variably
saturated geologic media: Applications to geothermal injectivity and CO2
geological sequestration. Comput Geosci-Uk 32, 145-165.
Zerai, B., Saylor, B.Z., Matisoff, G., 2006. Computer simulation of CO2 trapped through
mineral precipitation in the Rose Run Sandstone, Ohio. Appl Geochem 21, 223-240.
Zhang, L., Parthasarathy, H., Karamalidis, A., 2016. Investigation on arsenopyrite
dissolution and As (III) migration under geologic carbon storage conditions: A
numerical simulation approach. Greenhouse Gases: Science and Technology.
77
Zheng, L., Spycher, N., Varadharajan, C., Tinnacher, R.M., Pugh, J.D., Bianchi, M.,
Birkholzer, J., Nico, P.S., Trautz, R.C., 2015. On the mobilization of metals by CO2
leakage into shallow aquifers: Exploring release mechanisms by modeling field and
laboratory experiments. Greenhouse Gases: Science and Technology 5, 403-418.
Zheng, L.G., Apps, J.A., Zhang, Y.Q., Xu, T.F., Birkholzer, J.T., 2009. On mobilization of
lead and arsenic in groundwater in response to CO2 leakage from deep geological
storage. Chemical Geology 268, 281-297.
78
국문요약
지구온난화의 주요인으로 CO2가 지목 받는 가운데, 이를 저감하기 위한
대책의 일환으로 Carbon Capture & Storage (CCS, 이산화탄소 지중저장)가 각
광받고 있다. 안정성 있는 CCS를 위해서 그 위해성에 대한 연구 및 평가가 선
행 되어, 위기 발생 시 신속한 대처 및 저감 계획이 수립되어야 한다. 따라서
본 연구는 2D 반응성 거동 모델링을 활용하여 지중 주입된 CO2가 누출되 천
부 대수층으로 유입될 경우 발생 가능한 위해성을 규명하고 평가 하는 것을
목표로 한다.
누출된 CO2는 천부 대수층 내에서 플룸을 형성하고 내부의 pH를 저하시
킨다. 이는 대수층 매질을 구성하고 있는 각종 광물을 용해시키면서 수리전도
도를 증가 시킨다. 또 이 과정에서 광물 내 인체에 유해한 중금속 혹은 방사
성 핵 종이 존재할 경우 지하수를 오염시킨다. 본 연구에서는 황비철석
(Arsenopyrite)이 대수층 내 존재 할 경우 CO2 누출 시 발생하는 비소 오염에
대해 규명하고, 그 오염된 물을 양수하여 사용할 경우 발생하는 건강 위해성
에 대한 평가를 수행하였다. 끝으로 비소 오염 발생 시 위해성을 더 증가 시
키는 조건을 찾기 위해 민감도 분석 또한 수행하여 누출된 CO2 플룸의 거동
및 특징에 의해 위해성이 크게 영향 받는다는 사실을 밝혔다.