International Journal of Greenhouse Gas Control 102 (2020) 103156

Contents lists available at ScienceDirect

International Journal of Greenhouse Gas Control

journal homepage: www.elsevier.com/locate/ijggc

Wellbore injectivity response to step-rate CO2 injection: Coupled

thermo-poro-elastic analysis in a vertically heterogeneous formation
Hojung Jung a, *, D. Nicolas Espinoza a, Seyyed A. Hosseini b
a
Department of Petroleum and Geosystems Engineering, The University of Texas at Austin, USA
b
Gulf Coast Carbon Center, Bureau of Economic Geology, The University of Texas at Austin, Austin, Texas, USA

A R T I C L E I N F O A B S T R A C T

Keywords: Safe and permanent carbon geological storages require an elaborate analysis of geomechanical stability. Sub­
Carbon geological storage surface injection of CO2 changes the local temperature and pore pressure and, further, alters the stress due to
Cranfield thermo-poro-elastic responses.
Fracturing
A permanent CO2 storage testing was conducted in the water leg of Cranfield reservoir in Mississippi, USA,
Well testing
CO2injection
where hosted CO2 enhanced-oil recovery activities. During CO2 injection, the injection rate was ramped up twice
but the bottom-hole pressure did not increase with the imposed injection rates as expected. This unexpected field
observation suggests the possibility of an open-mode fracture development at the injector. However, the injector
response and potential fracture development have not been rigorously interpreted with detailed near-wellbore
temperature and pore pressure change upon the CO2 injection.
In this study, we performed history matching of CO2 injection using coupled thermo-poro-elastic reservoir
simulation to examine the possibility of fracturing. We built a reservoir model including vertical heterogeneity in
both petrophysical and geomechanical properties estimated from well-logging analysis and laboratory
experiments.
The simulation results show that CO2 injection changes stresses and support the hypothesis of development of
an open-mode fracture at the injector during the Cranfield test. The near-injector region exhibits a large tem­
perature reduction up to 55 ◦ C with ensuing effective horizontal stress reduction up to 9.1 MPa. However, a
caprock integrity issue is unlikely because of the horizontal stress contrast within layers and low hydraulic
communication with the injection zone.
This study indicates that the injectant temperature should be considered in the design of high-rate CO2
injector.

1. Introduction
Geological sequestration of carbon dioxide (CO2) can alleviate po­
tential carbon emissions to the atmosphere by injecting CO2 into
depleted reservoirs or brine formations (IPCC, 2005; Benson and Surles,
2006; Benson and Cole, 2008). However, injecting large amounts of CO2
at high rates into a formation may disturb the geomechanical equilib­
rium of the formation and lead to fault shear reactivation or open-mode
fractures (Rinaldi and Rutqvist, 2013; Bauer et al., 2016; Rutqvist et al.,
2016; Jung et al., 2018). Predicting the evolution of the stress state upon
CO2 injection is a complex problem that includes thermo-elastic,
poro-elastic, and chemo-elastic coupled processes (Espinoza et al.,
2011 & 2018; Zoback and Gorelick, 2012; Kim and Hosseini, 2013 &
2017). CO2 injection builds up reservoir pressure and alters local stress
depending on formation capacity, compressibility, permeability, and
connectivity to aquifer (Ehlig-Economides and Economides, 2010; Jung
and Espinoza, 2017& 2018). Further, injecting fluids at ambient tem­
perature (on surface) into a reservoir at high temperature results in rock
shrinkage and effective stress reduction (Luo and Bryant, 2011; Gor and
Prevost, 2013; Roy et al., 2018). CO2 injection also acidifies the for­
mation brine, and that may induce mineral dissolution and lower the
rock strength (Hangx et al., 2012; Aman et al., 2017; Jung and Espinoza,
2017; Espinoza et al., 2018). These mechanisms may alter the geo­
mechanical equilibrium and induce inelastic strains in the reservoir.
The Cranfield reservoir in Mississippi, USA implemented CO2 injec­
tion for enhanced oil recovery and carbon sequestration (Hovorka et al.,

* Corresponding author at: 200 E Dean Keeton, Austin, 78712, USA.

E-mail address: hjung@utexas.edu (H. Jung).

https://doi.org/10.1016/j.ijggc.2020.103156
Received 20 May 2018; Received in revised form 1 September 2020; Accepted 2 September 2020
Available online 2 October 2020
1750-5836/© 2020 Elsevier Ltd. All rights reserved.
H. Jung et al.
Fig. 1. Schematic diagram of Cranfield water leg Detailed Area of Study (DAS) for permanent storage of CO2.
Fig. 2. Imposed injection rate (a), bottom-hole pressure (BHP) (b), and BHP vs injection rate (c) at well CFU31F-1 and the expected simulation result with constant
permeability without fracture opening. Fig. 2 (c) shows the deviation of field data from simulation linear line due to possible fracture opening.
2013). A total of 0.5 million tons of CO2 were injected in the water leg
solely for carbon sequestration (Southeast Regional Carbon Sequestra­
tion Partnership – www.secarbon.org). We refer to this section of the
formation as Cranfield “Detailed Area of Study (DAS)” in this paper
(Fig. 1). The DAS area includes one injector (CFU31F-1) and two
observation wells (CFU31F-2 and CFU31F-3) perforated at the interval
of the Tuscaloosa sandstone (Hovorka et al., 2013; Lu et al., 2012).
During the injection in the DAS, the CO2 injection rate was initially 175
kg/min and then ramped up twice, from 175 to 300 kg/min and then
from 300 to 500 kg/min (Soltanian et al., 2016). Even though the in­
jection rate was nearly doubled during the second injection rate change,
the injection well did not experience an increase in bottom-hole pressure
(BHP), as expected from typical step-rate tests and confirmed by reser­
voir simulation (Fig. 2 & Kim and Hosseini, 2013; Soltanian et al., 2016).
The field observations of BHP suggest the development of an open-mode
fracture, after injection rate was ramped to 500 kg/min (Kim and Hos­
seini, 2013; Soltanian et al., 2016) and permeability modification
(Delshad et al., 2013; Min et al., 2017) in the near-wellbore region
during injection. Even though previous studies (Delshad et al., 2013;
Min et al., 2017) utilized permeability modification factors to match the
unexpected pressure response, the reasons for such permeability modi­
fication have not been studied in depth with geomechanical issue.
In this study, we built a reservoir model to simulate CO2 injection in
the DAS using CMG-GEM (Computer Modeling Group Ltd., 2013)
coupled with thermo-poro-elasticity and compositional fluid behavior.
The model uses petrophysical and geomechanical properties from
well-logging analysis calibrated with laboratory measurements using
field cores. Then we conducted history matching of the pressure
2
International Journal of Greenhouse Gas Control 102 (2020) 103156
response at the injector (CFU31F-1) without thermo-elasticity first using
the reservoir model. We expanded the model with thermos-elasticity
and analyzed the effects of thermally induced stress relaxation on
effective stress and the possibility of propagation of open-mode fractures
at the injector. Finally, we performed sensitivity analysis to injection
temperature, thermal expansion coefficient, and maximum fracture
permeability. The sensitivity analysis helps understand the effects of
various thermo-elastic parameters on the geomechanical stability of the
CO2 reservoir.
2. Cranfield DAS reservoir properties and model
Cranfield DAS model includes vertical heterogeneity in petrophysical
and geomechanical properties from well-logging analysis and laboratory
core measurements on Tuscaloosa sandstone. Detailed model construc­
tion is described in the following subsections.
2.1. Petrophysical properties of lower Tuscaloosa sandstone
We estimated reservoir properties from well-logging data analysis
from the injection (CFU31F-1) and one of the observation wells
(CFU31F-2). The distance between these two wells is 68 m (223 ft). The
two wells show similar rock types and sequences with slightly shifted
depth as a result of reservoir dip of 2 – 3◦ (Lu et al., 2013). Since the
purpose of this paper is to perform a geomechanical analysis of the
near-injector region, the simulation includes heterogeneity in the ver­
tical direction only for both petrophysical and geomechanical
properties.
H. Jung et al. International Journal of Greenhouse Gas Control 102 (2020) 103156

Fig. 3. Geological context of Cranfield DAS: (a) gamma ray, (b) compressional wave velocity, (c) porosity, (d) horizontal permeability, (e) Poisson’s ratio, (f) dy­
namic Young’s modulus, (g) static Young’s modulus, (h) pore compressibility, and (i) Biot coefficient. Red lines are well-logging analysis results, and dark blue lines
are the averaged values for simulation at injection zone.

Table 1
Information of samples from laboratory experiments.
Sample number Depth [m] Diameter [in] Length [in] Weight [g] Bulk density [g/cm3] Porosity Rock density [g/cm3] Permeability [mD]

38V 3189.9 0.97 2.00 44.30 1.82 0.31 2.13 10 - 20

30V 3192.9 0.98 1.74 42.30 1.96 0.26 2.22 20 - 40
26H 3188.6 0.97 1.66 36.97 1.83 0.31 2.14 150 - 250

Fig. 3 summarizes the well log analysis results and averaged prop­
Table 2
erties for the simulation model. We averaged property values every 1.1
Rock types applied to Frio reservoir modeling for capillary pressure and relative
m (3.6 ft) to balance our DAS model simulation computational cost and
permeability. (Assumed parameter includes Sm = 1).
the level of detail of the reservoir sequence and heterogeneity (Fig. 1).
Rock type E (GPa) v Cp (1/psi × 10− 6) α ϕavg

2.1.1. Porosity and permeability 1 13.71 0.255 2.35 0.824 0.177

We used clay corrected sand porosity (Fig. 3-c - Torres-Verdin, 2 13.45 0.240 2.40 0.837 0.181
2016). As discussed above, the injection and observation wells show 3 12.15 0.285 2.19 0.822 0.200
4 9.84 0.286 2.24 0.855 0.241
similar rock types and sequences. Hence, we adopted permeability
5 9.02 0.286 2.27 0.868 0.260
values estimated for the observation well CFU31F-2 after correcting for 6 15.23 0.324 2.04 0.728 0.153
the depth difference. The ratio between vertical and horizontal perme­ 7 8.68 0.304 2.17 0.861 0.269
ability is set to be 0.1 (kv/kh = 0.1), which is an average value from 8 (Caprock) 16.80 0.150 7.78 0.849 0.050
laboratory measurements using cores from the field (Table 1). 9 (Underlying layer) 25.20 0.150 25.9 0.774 0.010

2.1.2. Capillary pressure and relative permeability types based on porosity (including seven for the injection zone and two
We assigned heterogeneity to relative permeability and capillary for the caprock and underlying layers) in order to build an accurate
pressure distinguishing the caprock and the injection zone (sandstone) mechanical reservoir model (Table 2).
by adopting the corresponding parameters from the Brooks-Corey model
(Hosseini et al., 2013). 2.2.2. Unloading formation compressibility
Pore compressibility results in significant impact to local pore pres­
sure calculation upon injection (Jung and Espinoza, 2017). The model
2.2. Geomechanical properties
adopts multiple pore compressibility values according to the corre­
sponding rock types. We calculated the pore compressibility Cpore based
2.2.1. Dynamic and static elastic moduli
on the relationship between the unloading constrained elastic modulus
We calculated elastic moduli using sonic travel time. Since the shear (1− v)E
wave travel time was not measured at the injector, we applied the ratio M = (1+v)(1− 2v) and porosity ϕ:
between compressional wave velocity VP and shear wave velocity VS
ϕ
from the observation well (CFU31F-2) to calculate VS at the injector. The Cpore ≈ (1)
M
ratio VP/VS varies from 1.5 to 1.7 at the injection zone, slightly lower
than the assumed value 2.0 in Carter and Spikes (2013) and Daley et al. The calculated unloading formation compressibility is assigned to the
(2014). The calculated static moduli were calibrated with the laboratory nine rock types (Table 2 and Fig. 3-h).
measured ratio between static and dynamic moduli approximately equal
to 0.56 (Jung and Espinoza, 2017).
We assigned the calculated static elastic moduli to nine different rock

3
H. Jung et al. International Journal of Greenhouse Gas Control 102 (2020) 103156

injection cycles. The reservoir domain size is 4.0 × 4.0 km2 based on
sensitivity analysis. The simulation domain thickness is 76.8 m
including both injection zone and parts of caprock and underlying
layers.

2.3.3. Geomechanical boundary conditions

We assumed zero-lateral strain at boundaries and constant vertical
stress conditions above the caprock. We applied total vertical stress Sv =
66.2 MPa at 3150 m of depth (top of the caprock layer in the simulation
domain - stress gradient of 21 MPa/km (0.933 psi/ft) (Fig. 5). The initial
stress condition of the numerical model is shown in Fig. 5. Using the
vertical stress applied for simulation, we also calculated horizontal
stress assuming a stacked poroelastic medium without tectonic strains:
( v ) ( )
1 − 2v
Sh = Sv + αPp (3)
Fig. 4. Reservoir domain and simulation boundary conditions. 1− v 1− v

where Sv and Sh are vertical and minimum horizontal total stresses, Pp is

2.2.3. Biot coefficient
pore pressure, v is Poisson’s ratio, and α is the Biot coefficient (Lorenz
The DAS model uses various Biot coefficient (α) as shown in Fig. 3-i.
et al., 1991). The initial stress condition of the numerical model is close
Biot coefficient is calculated with the following equation:
to the analytical solution and captures the local minimum horizontal
α=1−
K
(2) stress at ~ 3186 m (Fig. 5). We assumed the initial maximum total
Kunj horizontal stress is 6.9 MPa (1000 psi) higher than the minimum hori­
zontal stress to avoid the effects of possible principal stress rotation near
where K is bulk drained modulus and Kunj is the unjacketed bulk the fracture region.
modulus. The parameter Kunj is estimated to be 53 GPa using experi­
mentally measured poroelastic response (cyclic pore-pressure and 2.3.4. Fracture permeability model
confining pressure loading – Bouteca and Gueguen, 1999) of Tuscaloosa We adopted dual permeability and the Barton-Bandis model (Barton
sandstone samples from the observation well CFU31F-3 (Jung and et al., 1985) to capture fracture opening and the permeability evolution.
Espinoza, 2017). Considering Biot coefficient not equal to one yields The Barton-Bandis model enables permeability changes depending on
appropriate initial fracture gradient and effective stress alteration upon effective stress. We assigned dual permeability elements aligned with
pore pressure change (Kim and Hosseini, 2017). the direction of maximum horizontal total stress SHmax. The fracture
permeability is the same as the matrix permeability initially. Once the
Biot effective stress reaches zero, the fracture activates, and the grid
2.3. Cranfield simulation domain and boundary conditions
block permeability switches to a fractured grid block permeability (kf, =
100 D, assumed in this study). The fracture permeability (while it closes)
2.3.1. Simulation domain
is computed by the following equations:
The DAS simulation includes 87,261 (59 × 51 × 29) degrees of
freedom with refined mesh near the injector region (Fig. 4). The vertical
grid block size is 1.1 m. Horizontal grid blocks are composed of thirteen
6 m-length blocks, one of 12 m, two of 40 m, and six of 100 m from the
injector in each x and y direction, where non-zero CO2 saturation values
are observed (Fig. 8). Outside of this region, gridding includes incre­
mental grid block sizes up to 200 m. A tensor product refinement was
used to better capture changes in the DAS, where the injection well is
located. This approach is especially useful when a limited number of
reservoir properties is available from well logs, seismic observations and
geological models. Result grid independence was checked by refining
grid blocks, with maximum grid block size as large as 40 m where the
CO2 plume is located. The simulation results with refined meshing show
less than 2 % of error in CO2 plume size. More details about simulations
with finer grid blocks are described in Appendix. A. The compositional
flow model in CMG-GEM uses the Peng-Robinson cubic equation of state
(PR-EOS) for describing fluid phase behavior for brine and CO2
(Table 3). Henry’s law calculates the CO2 solubility as 1.2 (mol of CO2)/
(kg of water) using the coefficients in Table 3 at P = 31.0 MPa and T =
127 ◦ C.

2.3.2. Flow boundary conditions and the domain model size

Fig. 5. Initial stress conditions in the injection zone from analytical solutions
We assumed no flow boundary condition by increasing the model
and numerical simulation.
size large enough to minimize the boundary effects during the CO2

Table 3
EOS parameters for CO2 solubility calculation.
Tc [R◦ ] Pc [psi] Zc ω[-] Mw [g/mol] P [-] Henry’s law constant [kPa] Pref for Henry’s law [MPa] Partial molar volume [l/mol]

CO2 547.56 1070.4 0.302 0.224 44.01 78.0 749,588 31.0 0.037

4
H. Jung et al. International Journal of Greenhouse Gas Control 102 (2020) 103156

Table 4 temperature dropped to 72 ◦ C during injection according to field mea­

Barton-Bandis parameter for fracture permeability evolution. surements (Kim and Hosseini, 2013). We set the initial temperature of
Initial fracture aperture [mm] 0.006 the reservoir and injected CO2 accordingly. The simulation adopts a
linear thermal expansion coefficient αT = 1.3 × 10− 5 1/◦ C in the range
Initial normal fracture stiffness [kPa/m] 1.5 × 109
of sandstones (1.3 ~ 2.0 × 10− 5 1/◦ C - Fjaer et al., 2008).
Fracture opening stress [kPa] 0
Hydraulic fracture permeability [mD] 100,000
Fracture closure permeability [mD] 233 2.4. Description of performed simulations
Residual value of fracture closure permeability [mD] 1
Reservoir simulation included various cases (Table 5):
( )4
kf = kccf
e
≥ krcf (4) • Baseline scenario simulation without thermo-elasticity (BC wo TH),
eo • Thermo-elastic history match simulation with thermo-elsaticity (BC
TH),
σ ’n • Sensitivity analysis cases varying injection temperature (IT), thermal
e = e0 − / (5)
kni + σ ’n Vm expansion coefficient (THexp), and fracture permeability (FP).

where, kccf is fracture closure permeability at zero effective stress (at the
onset of fracture closure), krcf is minimum residual permeability of
closed fracture, e is fracture aperture, e0 is initial fracture aperture, kni is
the initial normal fracture stiffness, Vm is maximum fracture closure. If
the fracture block is activated, counting the activated blocks renders the
fracture propagation length.
Since the Barton-Bandis model follows a joint-system concept, we
adjusted the maximum activated fracture permeability as 100 D equiv­
alent to approximately 1.5 mm aperture of each fracture. Each fracture
is spaced every 3.0 m and the direction of fracture propagation is
perpendicular to minimum principal stress direction. The details of pa­
rameters used for Barton-Bandis model are available in Table 4. The
fracture model in CMG-GEM applies the Biot effective stress for Barton-
Bandis fracture permeability calculation. This feature delays a fracture
opening around 0.1 days in the model compared to applying Terzaghi’s
effective stress because Biot coefficient is less than one.
First, we conducted a simulation without thermo-elasticity. Injection
schedule is adjusted to mimic the field injection schedule (Fig. 6). We
performed history matching of the DAS pressure transient by changing
fracture permeability and thermal expansion coefficient and including
thermo-elasticity as described in Section 2.3. The objective of CO2 in­
jection simulations was to quantify the stress alteration from CO2 in­
jection and the effect of thermo-elasticity. Further, the simulation
examines the possibility of open-mode failure at the injection zone
during CO2 injection. Based on the history matched case, we conducted
sensitivity analysis on injection temperature, thermal expansion coeffi­
cient, and fracture permeability. Both the injection temperature and the
thermal expansion coefficient determine the amount of stress reduction
as the rock temperature decresases from the near-injector region. Frac­
ture permeability affects the BHP when the fracture opens. The details of
simulations inputs are available in Table 5.

2.3.5. Thermo-elastic properties and initial conditions

The initial bottom-hole temperature at CFU31F-1 is 127 ◦ C. The

Table 5
Simulation input values (injection temperature, thermal expansion coefficient and fracture permeability) varied in numerical simulations.
BC wo TH BC TH IT1 IT2 IT3 IT4 FP1 FP2 THexp1 THexp2
Thermo-elasticity Isothermal Non-Isothermal Non-isothermal

Tinjection [◦ C] 72 72 48 60 85 97 72 72 72 72
kfracture [D] 100 100 100 100 100 100 10 1000 100 100
αT [1/◦ C × 106] 13 13 13 13 13 13 13 13 9.9 16

Fig. 6. Injection schedule and BHP of CFU31F-1: field data, simulation without thermo-poro-elasticity, and simulation with thermo-poro-elasticity.

5
H. Jung et al.
Fig. 7. Biot effective stress at the wellbore block at depth 3185 m (local the
least horizontal minimum stress) with and without thermo-elasticity.
3. Results and discussion
3.1. History-match simulation without thermo-elasticity
The simulation input properties and boundary conditions are
Fig. 8. CO2 Saturation and temperature in the DAS at the end of simulation (220 days after the initiation of the injection). Cross section perpendicular to Shmin. The
wellbore CFU31F-1 is in the center.
6
International Journal of Greenhouse Gas Control 102 (2020) 103156
described in detail in section 2.3. The results show that BHP is directly
proportional to injection rate, and pressure never reaches the minimum
horizontal stress (updated with poro-elastic effects) (Fig. 5 & 6). The
effective stress is positive throughout the three injection cycles, so that,
fractures never initiate in the Barton-Bandis fracture model (Fig. 7).
Hence, the simulation without thermo-elasticity fails to predict the
relatively flat pressure response with increasing injection rate observed
in the field.
3.2. History-match simulation with thermo-elasticity
The simulation with thermo-elasticity shows a drastic reduction in
minimum horizontal stress (Fig. 6), and the Biot effective stress reaches
zero at the depths of 3186 m (local minimum of Shmin) after 31 days of
CO2 injection when the first jump in injection rate happens. “Negative”
effective stress activates block fracture permeability and increases its
value. The fracture half-length was 3.0 m at 31 days and increased up to
51.8 m at the end of the simulation (232 days). Comparing the simula­
tion without thermo-elasticity to the one with thermo-elasticity, we
observe little difference in pressure response before 166 days. This is
because the fractured blocks in the thermo-elastic simulation extend less
than 33.5 m from the wellbore before the second injection rate increase.
Injection of CO2 pumped from bottom-hole temperature conditions
(T =72 ◦ C and Treservoir = 127 ◦ C) results in a change of total minimum
horizontal stress of ~ 9.7 MPa (1406 psi) and Biot effective horizontal
H. Jung et al.
Fig. 9. Sensitivity analysis results of (a) Temperature difference between
injected CO2 and the formation, (b) Thermal expansion coefficient, and (c)
Maximum opened fracture permeability.
stress of ~ 9.0 MPa (1305 psi) in the simulation (Fig. 7). The steady-state
analytical solution for change of temperature everywhere in the for­
mation, one-dimensional vertical strain, constant vertical stress, and no
change of pore pressure yields a horizontal stress reduction equal to
[ ] [◦ ] [ ]
13∙10− 6 ◦1C ∙55 C ∙12.14 GPa
αT ∆TE
∆Shmin = = = 12.1MPa (6)
1− v 1 − 0.285
The numerical solution agrees reasonably well with the one-
dimensional strain assumption in the steady-state analytical solution
considering the transient process of temperature change with CO2 in­
jection and fracture propagation in simulation. Even though the CO2
plume extends up to 400 m, the reduction in temperature is significant
only around 65 m away from the wellbore (Fig. 8). Therefore, horizontal
stress alteration due to thermal stress relaxation is more pronounced in
the near-wellbore region.
7
International Journal of Greenhouse Gas Control 102 (2020) 103156
3.3. Sensitivity study to injection temperature
Injectant temperature has a direct impact on the change of total
horizontal stress. Low injection temperature reduces the horizontal total
stress at the near-wellbore region significantly (Fig. 9). However, the
injection temperature does not change the BHP directly, rather it
changes the fracture propagation length and the resulting BHP (fixed
injection rate). Before the large fracture opening and propagation at the
second injection rate increase at 172 days, the BHP did not show a
significant difference among simulations IT1 to IT4. After the develop­
ment of a large open-mode fracture in simulation at 172 days, the BHP
begins to show notable gaps between the two simulations IT1 and IT4.
The simulation case with the highest injection temperature T =98 ◦ C
(ΔT =29 ◦ C) shows the shortest fracture propagation length equal to
39.6 m at the end of the simulation. The simulation with the lowest
injection temperature T =48 ◦ C (ΔT =79 ◦ C) shows fracture propagation
up to 70.1 m. Further, the difference of Biot effective stress between two
cases is higher than 9 MPa, equivalent to twice the stress change
observed in the simulation without thermo-elasticity. These simulations
clearly demonstrate that the temperature of the injected fluid alters the
horizontal total stress and impacts to fracture opening and propagation,
and the resulting BHP. Therefore, the temperature of the injected fluid
should be carefully analyzed for CO2 storage injection to avert devel­
opment of open-mode fractures.
3.4. Sensitivity study to thermal expansion coefficient
The thermal expansion coefficient αT determines horizontal stress
reduction from temperature change. The simulation using the highest
thermal expansion coefficient αT = 1.60 × 10− 5 1/◦ C (THexp2) results in
earlier fracture opening and longer fracture propagation (58.0 m half
length) than the case with lowest thermal expansion coefficient αT =
0.99 × 10− 5 1/◦ C (THexp1) with 39.6 m of fracture half-length. Simu­
lation THexp2 shows 9 MPa decrease in Biot effective stress compared to
4 MPa decrease of the stress in Simulation THexp1 (Fig. 9-b). The dif­
ference in Biot effective stress between the two simulations is equivalent
to half of the stress change in the simulation without thermo-elasticity.
The range of thermal expansion coefficient in sandstones is 1.3–2.0 ×
10− 5 1/◦ C (Fjaer et al., 2008). Sensitivity analysis to elastic moduli
yields a linear relationship between the Biot effective stress alteration
according to Eq. 6. Fig. 11 shows an upper bound of the decrease of
minimum principal stress induced by thermo-elastic effect with different
combinations of thermal expansion coefficient and Young’s modulus
using Eq. 6. Hence, accurate determination of the rock mass Young’s
modulus and the thermal expansion coefficient greatly contributes to
predicting the fracture initiation time, length, and the resulting BHP
response.
3.5. BHP sensitivity to fracture permeability
Fracture opening accelerates the fluid pressure diffusion by
increasing the effective permeability of the reservoir. Therefore, the
amount of pressure decrease depends on the transmissibility of fractures.
Our sensitivity analysis predicts the lowest BHP (P = 40.8 MPa at 225
days) for fracture permeability kf = 1000 D (Simulation FP2). The BHP is
1.43 MPa higher for Simulation FP1 with kf = 10 D. Fracture half-length
at the end of the simulation is 45 m and 64 m for FP2 and FP1 respec­
tively. There are no significant differences in Biot effective stress be­
tween these two simulations (Fig. 9-c).
3.6. Horizontal stress reduction due to chemically-induced creep from
CO2-acidified brine
We conducted laboratory experiments to observe the effect of CO2-
acidified brine injection on stress alteration (Jung and Espinoza, 2017,
Fig. 10-a) triggered by rock deformation and strength alteration (Hangx
H. Jung et al.
Fig. 10. (a) Stress and pressure signals during injection of CO2-acidified brine
in a core sample of Tuscaloosa sandstone. (b) Volumetric strain changes upon
injection of CO2-acidified brine. The mean effective stress is constant σmean =
18.0 ± 0.13 MPa during injection.
Fig. 11. Upper bound estimation of minimum principal stress change as a
function of temperature reduction due to CO2 injection injection based on Eq. 6
assuming ν (0.285), αT (1.3⋅10− 6 1/◦ C), and E (12.14 GPa). The figure shows
two CO2 storage field examples in which there was a significant temperature
difference between injected CO2 and the reservoir: Cranfield, Mississippi
(Hovorka et al., 2013; Kim and Hosseini, 2013) and Snohvit in Norway (Hansen
et al., 2013; Chiaramonte et al., 2014). The 50 % reduction reflects uncertainty
in the rock mass stiffness and transient effects during injection.
et al., 2012; Espinoza et al., 2018). Before the CO2-acidified brine in­
jection, the sample was saturated with brine at a pore pressure of 2.1
MPa and constant total stresses Saxial = 22.8 MPa and Sradial = 11.7 MPa
for an hour until creep was negligible. Then we injected the CO2-ac­
idified brine into the sample (Fig. 10-a) Fig. 10-b shows volumetric
strain changes with four cycles of CO2-acidified brine injection, aver­
agely 60 pore volumes of injection at each cycle. The first injection cycle
exhibits the highest strain change rate, and the total change of volu­
metric strain is 2.9⋅10− 3 during four cycles (Fig. 10-b).
The results suggest that injection of CO2-acidified brine increases the
creep rate quickly after contacting with the rock. Other experiments
have suggested rapid responses of chlorite clays to CO2-acidified brine
and disintegration in Tuscaloosa sandstone and Mt. Simon sandstone
(Yoksoulian et al., 2013; Rinehart et al., 2016; Roy et al., 2018) with
8
International Journal of Greenhouse Gas Control 102 (2020) 103156
negligible chlorite dissolution in CO2-acidified brine (Lu et al., 2012;
Islam et al., 2016). Hence, it seems likely that clay swelling, debonding,
and dislodging -by changes in pore fluid chemistry- may have contrib­
uted to the damage of chlorite cements that increased creep rates in our
experiments.
At large time-scales, interaction of CO2-acidified brine and rock
minerals could lead to vertical deformation and horizontal stress
relaxation from enhanced creep (Espinoza et al., 2018). For example, a
change of strain of ~ 1×10− 3 (change of radial strain in horizontal
sample 26 H) multiplied by the height of the injection interval (20 m)
results in an upper bound estimation of a vertical displacement (2 cm). A
change of 2 cm at 3 km of depth is negligible regarding subsidence. On
the other hand, a change of lateral stress under zero lateral strain caused
by chemical-stress-relaxation is the product of chemically-induced strain
(~0.3×10− 3 from the change of axial strain in horizontal sample 26 H)
and the plane strain modulus. Hence, the horizontal stress reduction
could be as much as Δσhmin = ε × (1− Eν2 ) = 0.3×10− 3 × 10 GPa /
(1-0.32) = 3.3 MPa (Shovkun and Espinoza, 2018).
4. Conclusions
We performed a thermo-poro-elastic simulation of CO2 injection in
the water leg of the Cranfield reservoir and additional laboratory ex­
periments in order to investigate the geomechanical effects on wellbore
response. Simulation results and sensitivity analyses show that:
• Heterogeneity in geomechanical properties such as Young’s
modulus, Poisson’s ratio, and pore compressibility is critical to pre­
dicting local pore pressure buildup, horizontal stress and the locali­
zation of potential open-mode fractures in the wellbore.
• Injection of CO2 from surface ambient conditions lowers temperature
of the reservoir rock and leads to horizontal stress reduction in a
near-wellbore region. The simulation results support the hypothesis
of development of an open-mode fracture at the injector in the
Cranfield CO2 injection site.
• Despite the occurrence of an open mode fracture, the model also
predicts fracture containment without propagation into bounding
sealing layers.
• Sensitivity analysis of injection temperature and thermal expansion
coefficient shows significant impact of these parameters on hori­
zontal stress and fracture propagation. By lowering the injectant
temperature approximately 45 %, we observed Biot effective stress
decreases equivalent to four times of the stress change observed in
simulation without thermo-elasticity. Changing 23 % in thermal
expansion coefficient results in change of Biot effective stress,
equivalent to 50 % of the stress change without thermo-elasticity.
The reduction in fracture propagation pressure is also proportional
to the Young’s modulus. Therefore, accurate calculation of rock
thermal and mechanical properties is critical to avoid fracture
opening and propagation of large fractures.
• Laboratory experiments demonstrate that CO2-acidified brine in­
duces enhanced creep in chlorite-rich Tuscaloosa Sandstone. How­
ever, the impact of chemical effects at the length scale of a reservoir
and time scale of injection is limited compared to thermal effects on
stress reduction.
CRediT authorship contribution statement
Hojung Jung: Investigation, Methodology, Software, Visualization,
Writing - original draft. D. Nicolas Espinoza: Funding acquisition,
Writing - review & editing, Supervision. Seyyed A. Hosseini: Resources.
Declaration of Competing Interest
The authors declare that they have no known competing financial
H. Jung et al. International Journal of Greenhouse Gas Control 102 (2020) 103156

interests or personal relationships that could have appeared to influence
the work reported in this paper.
Acknowledgements
This work was supported as part of the Center for Frontiers of Sub­
surface Energy Security, an Energy Frontier Research Center funded by
the U.S. Department of Energy (DOE), Office of Science, Basic Energy
Sciences under Award DE-SC0001114 and National Energy Technology
Laboratory (DOE-NETL) through grant number DE-FE0023314.
Appendix A
We conducted sensitivity study of grid refinement. We performed
two different refinements in horizontal direction. Refined case 1 and 2
both are refined until 367 m from the injector, where the length of the
CO2 plume is observed at the end of the simulation. Horizontal grid
refinements from the injector for both x and y directions are composed
of blocks as follows:
• Base case: thirteen blocks of 6 m (covering the distance from the
injector to an observation well), one block of 12 m, two blocks of 40
m, and two blocks of 100 m.
• Refined case 1: thirteen blocks of 6 m, one block of 12 m, and seven
blocks of 40 m.
• Refined case 2: thirteen blocks of 6 m, two blocks of 12 m, three
blocks of 22.67 m, and five blocks of 40 m.
Fig. A1 shows the simulation results of CO2 saturation and formation
temperature at the end of simulation – 220 days of CO2 injection. The
results show similar sizes of the CO2 plume with 3.5 % difference in
volumetric size with changing grid refinement (Table A). The CO2 plume
front at the layer with lowest initial Shmin (at the depth of 3185 m) is
comparable with 3.3 % difference (Table A). Temperature reduction and
the resulting stress alteration upon CO2 injection exhibit similar varia­
tions (Table A1 & Fig. A1). These results imply that the base case
simulation has enough refinement to capture thermo-mechanical effects
in the near wellbore region to support the conclusions of this study.

Fig. A1. CO2 Saturation and temperature in the DAS at the end of simulation with two different refined grid discretization cases (220 days after the initiation of the
injection). Cross section perpendicular to Shmin. The wellbore CFU31F-1 is in the center.

Table A1
Simulation results based on grid discretization at the depth of 3185m (the layer with lowest initial Shmin).
Base case Refined case 1 Refined case 2 Error based on Refined case 2 (%)
3 6 6 6
Total CO2 plume volumetric size (higher than 6% saturation) (m ) 2.28×10 2.22×10 2.21×10 3.5
CO2 plume front from the injector (m) 317.0 307.0 307.0 3.3
Thermal reduction front from the injector (higher than 5 ◦ C reduction) (m) 48.0 48.0 48.0 0.0
Temperature at the distance of 111 m, the observation well located (31F-3) (◦ C) 126.8 126.6 126.6 0.2
Shmin at the distance of 111 m, the observation well located (31F-3) (MPa) 40.3 40.5 40.5 0.5

Base case Refined case 1 Refined case 2.

9
H. Jung et al. International Journal of Greenhouse Gas Control 102 (2020) 103156

References Islam, A., Sun, A.Y., Lu, J., 2016. Simulating in-zone chemistry changes from injection
time to longer periods of CO2 storage. Environ. Earth Sci. 75.
Jung, H., Espinoza, D.N., 2017. Chemo-Poromechanical Properties of Tuscaloosa
Aman, M., Espinoza, D.N., Ilgen, A.G., Major, J.R., Eichhubl, P., Dewers, T.A., 2017. CO2-
Sandstone: Implications on CO2Geological Storage. 51st U.S. Rock Mechanics/
induced chemo-mechanical alteration in reservoir rocks assessed via batch reaction
Geomechanics Symposium, 25-28 June. ARMA, San Francisco, California, USA,
experiments and scratch testing. Greenhouse Gas Sci. Technol. 8, 133–149.
2017-0303.
Barton, N., Bandis, S., Bakhtar, K., 1985. Strength, deformation and conductivity
Jung, H., Singh, G., Espinoza, D.N., Wheeler, M.F., 2018. Quantification of a maximum
coupling of rock joints. Int. J. Rock Mech. Min. Sci. 22, 121–140.
injection volume of CO2to avert geomechanical perturbations using a compositional
Bauer, R.A., Carney, M., Finley, R.J., 2016. Overview of microseismic response to CO2
fluid flow reservoir simulator. Adv. Water Resour. 112, 160–169.
injection into the Mt.SImon saline reservoir at the Illinois Basin-Decatur Project. Int.
Kim, S.H., Hosseini, S.A., 2013. Above-zone pressure monitoring and geomechanical
J. Greenh. Gas Control. 54, 378–388.
analyses for a field-scale CO2 injection project in Cranfield, MS. Society of Chemical
Benson, S.M., Cole, D.R., 2008. CO2 sequestration in deep sedimentary formations.
Industry. Greenhouse Gas Sci Technol. 4, 81–98.
Elements 4, 325–331.
Kim, S.H., Hosseini, S.A., 2017. Study on the ratio of pore-pressure/stress changes during
Benson, S.M., Surles, T., 2006. Carbon dioxide capture and storage: an overview with
fluid injection and its implications for CO2 geologic storage. J. Pet. Sci. Eng. 149,
emphasis on capture and storage in deep geological formations. Proc. Ieee 94,
138–150.
1795–1805.
Lorenz, J.C., Teufel, L.W., Warpinski, N.R., 1991. Regional fractures; I, A mechanism for
Carter, R.W., Spikes, K.T., 2013. Sensitivity analysis of Tuscaloosa sandstones to CO2
the formation of regional fractures at depth in flat-lying reservoirs. Am. Associat.
saturation, Cranfield field, Cranfield, MS. Int. J. Greenh. Gas Control. 18, 485–496.
Petrol. Geol. 5, 1714–1737.
Chiaramonte, L., White, J.A., Tranior-Guitton, W., 2014. Probabilistic geomechanical
Lu, J., Kharaka, Y.K., Thordsen, J.J., Horita, J., Karamalidis, A., Griffith, C., Hakala, J.A.,
analysis of compartmentalization at the Snohvit CO2 sequestration project.
Ambats, G., Cole, D.R., Phelps, T.J., Manning, M.A., Cook, P.J., Hovorka, S.D., 2012.
J. Geophys. Res. Solid Earth 120, 1195–1209.
CO2–rock–brine interactions in Lower Tuscaloosa Formation at Cranfield CO2
Computer Modeling Group Ltd, 2013. Compositional & Unconventional Reservoir
sequestration site, Mississippi, U.S.A. Chem. Geol. 291, 269–277.
Simulation. Calgary.
Lu, J., Kordi, M., Hovorka, S.D., Meckel, T.A., Christopher, C.A., 2013. Reservoir
Daley, T.M., Henderickson, J., Queen, J.H., 2014. Monitoring CO2 storage at Cranfield,
characterization and complications for trapping mechanisms at Cranfield CO2
Mississippi with time-lapse offset VSP – using integration and modeling to reduce
injection site. Int. J. Greenh. Gas Control. 18, 361–374.
uncertainty, MS. Energy Procedia 63, 4240–4248.
Luo, Z., Bryant, S., 2011. Influence of Thermo-Elastic Stress on Fracture Initiation During
Delshad, M., Kong, X., Tavakoli, R., Hosseini, S.A., Wheeler, M.F., 2013. Modeling and
CO2Injection and Storage. Energy Procedia 4, 3714–3721.
simulation of carbon sequestration at Cranfield incorporating new physical models.
Min, B., Wheeler, M.F., Sun, A.Y., 2017. Parallel multiobjective optimization for the
Int. J. Greenh. Gas Control. 18, 463–473.
coupled Compositional/Geomechanical modeling of pulse testing. SPE Reservoir
Ehlig-Economides, C., Economides, M.J., 2010. Sequestering carbon dioxide in a closed
Simulation Conference 2017 at Mongomery 20–22. February.
underground volume. J. Pet. Sci. Eng. 70, 123–130.
Rinaldi, A.P., Rutqvist, J., 2013. Modeling of deep fracture zone opening and transient
Espinoza, D.N., Kim, S.H., Santamarina, J.C., 2011. CO2 geological storage -
ground surface uplift at KB-502 CO2 injection well, in Salah, Algeria. Int. J. Greenh.
Geotechnical implications. Ksce J. Civ. Eng. 15, 707–719.
Gas Control. 12, 155–167.
Espinoza, D.N., Jung, H., Major, J.R., Sun, Z., Ramos, M.J., Eichhubl, P., Balhoff, M.T.,
Rinehart, A.J., Dewers, T.A., Broome, S.T., Eichhubl, P., 2016. Effects of CO2 on
Choens, R.C., Dewers, T.A., 2018. CO2charged brines changed rock strength and
mechanical variability and constitutive behavior of the Lower Tuscaloosa Formation,
stiffness at Crystal Geyser, Utah: Implications for leaking subsurface CO2 storage
Cranfield Injection Site, USA. Int. J. Greenh. Gas Control. 53, 305–318.
reservoirs. Int. J. Greenh. Gas Control. 73, 16–28.
Roy, P., Morris Morris, J.P., Walsh, S.D., Iyer, J., Carroll, S., 2018. Effect of thermal stress
Fjaer, E., Holt, R.M., Raaen, A.M., Risnes, R., Horsrud, P., 2008. Petroleum Related Rock
on wellbore integrity during CO2 injection. Int. J. Greenh. Gas Control. 77, 14–26.
Mechanics, 2nd edition, Vol. 53. Elsevier Science.
Rutqvist, J., Rinaldi, A.P., Cappa, F., Jeanne, P., Mazzoldi, A., Urpi, L., Guglielmi, Y.,
Gor, G.Y., Prevost, J.H., 2013. Effect of CO2injection temperature on caprock stability.
Vilarrasa, V., 2016. Fault activation and induced seismicity in geological carbon
Energy Procedia 37, 3727–3732.
storage–Lessons learned from recent modeling studies. J. Rock Mech. Geotech. Eng.
Hangx, S., Linden, A., Marcelis, F., Bauer, A., 2012. The effect of CO2 on the mechanical
8, 789–804.
properties of the Captain Sandstone: geological storage of CO2 at the Goldeneye field
Shovkun, I., Espinoza, D.N., 2018. Geomechanical implications of dissolution of
(UK). Int. J. Greenh. Gas Control. 19, 609–619.
mineralized natural fractures in shale formations. J. Pet. Sci. Eng. 160, 555–564.
Hansen, O., Gilding, D., Nazarian, B., Osdal, B., Ringrose, P., Kristoffersen, J., Eiken, O.,
Soltanian, M.R., Amooie, M.A., Cole, D.R., Graham, D.E., Hosseini, S.A., Hovorka, S.,
Hansen, H., 2013. Snohvit: the history of injecting and storing 1 Mt CO2 in the fluvial
Pfiffner, S.M., Phelps, T.J., Moortgat, J., 2016. Simulating the Cranfield geological
Tubaen Fm. Green House gas Technology-11. Energy Procedia 37, 3565–3573.
carbon sequestration project with high-resolution static models and an accurate
Hosseini, S.A., Lashgari, H., Choi, J.W., Nicot, J., Lu, J., Hovorka, S.D., 2013. Static and
equation of state. Int. J. Greenh. Gas Control. 54, 282–296.
dynamic reservoir modeling for geological CO2 sequestration at Cranfield,
Torres-Verdin, C., 2016. Integrated Geological-Petrophysical Interpretation of Well Logs.
Mississippi, U.S.A. Int. J. Greenh. Gas Control. 18, 449–462.
Course Notes of Fundamentals of Well Logging. The University of Texas at Austin.
Hovorka, S.D., Meckel, T.A., Trevino, R.H., 2013. Monitoring a large-volume injection at
Yoksoulian, L.E., Freiburg, J.T., Butler, S.K., Berger, P.M., Roy, W.R., 2013.
Cranfield, Mississippi—project design and recommendations. Int. J. Greenh. Gas
Mineralogical alterations during laboratory-scale carbon sequestration experiments
Control. 18, 345–360.
for the Illinois Basin. Energy Procedia 37, 5601–5611.
IPCC, 2005. Underground geological storage. In: Metz, B., Davidson, O., Coninck, H.,
Zoback, M.D., Gorelick, S.M., 2012. Earthquake triggering and large-scale geologic
Loos, M., Meyer, L. (Eds.), IPCC Special Report on Carbon Dioxide Capture and
storage of carbon dioxide. Proc. Natl. Acad. Sci. U.S.A 109, 10164–10168.
Storage, Prepared by Working Group III of the Intergovernmental Panel on Climate
Change. Cambridge University Press, Cambridge, UK, and New York, USA,
pp. 195–276.

10