You are on page 1of 19

Earth-Science Reviews 239 (2023) 104369

Contents lists available at ScienceDirect

Earth-Science Reviews
journal homepage: www.elsevier.com/locate/earscirev

Seismicity induced by geological CO2 storage: A review


Yuxiang Cheng a, b, Wenna Liu a, Tianfu Xu a, Yanjun Zhang c, Xinwen Zhang a, d, Yunyan Xing e,
Bo Feng a, *, Yi Xia f
a
Key Lab of Groundwater Resource and Environment, Ministry of Education, Jilin University, Changchun, China
b
Engineering Research Center of Geothermal Resources Development Technology and Equipment, Ministry of Education, Jilin University, Changchun, 130026, China
c
Construction Engineering College, Jilin University, Ximinzhu Street, Changchun, China
d
Shuangyang District Procuratorate of Changchun City, Changchun, China
e
Jingyu County Development and Reform Bureau, Jingyu County, Bai Shan City, China
f
Northeast Electric Power Design Institute, Changchun, China

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

Keywords: Geological storage is a valuable strategy for reducing CO2 emissions to the atmosphere, although seismicity
Carbon capture and storage induced by CO2 injection can be a serious hazard that also becomes an obstacle to the development of CO2
Injection induced seismicity geological storage. The most important challenge in this field is fault systems that are difficult to detect and that
CO2 injection
have complex activation mechanisms, making the evaluation, prediction and control of CO2 injection induced
THMC coupling
Seismic magnitude prediction
seismicity extremely difficult. It is also challenging to determine the triggering mechanism of injection induced
seismicity. To promote the solution of these problems, we first review the experience and lessons learned from
recent induced seismicity monitored in CO2 geological storage (CGS) projects, and summarize the mechanisms
that can be used to analyze CO2 injection induced seismicity, including critical pressure theory, Biot’s incre­
mental strain theory, rate- and state-dependent frictional theory and fracture potential theory. We then discuss
the theory and modeling of thermo-hydro-mechanical-chemical (THMC) coupling in CGS-induced seismicity.
Knowledge of THMC coupling is an efficient way to improve the prediction of fault activation and seismic ac­
tivity and enables characterization of pore pressure perturbation, temperature changes, and stress and
geochemical effects. Through THMC simulation, we can more accurately characterize the change process of
stress field, analyze and speculate the triggering and spatio-temporal evolution of induced earthquakes. We also
summarize the advantages and disadvantages of maximum magnitude prediction based on statistical and
physical models. The statistical method is easy to use, but ignores the physical characteristics of the reservoir.
Although physical method overcomes this deficiency, obtaining sufficient modeling input parameters is always
an challenging work. Finally, we analyze the challenges involved in the seismicity forecasting, including the
quantification of stress state, the identification and characterization of complex fault system, the seismic miti­
gation injection scheme design, and reasonable seismic risk analysis model selection. This paper focuses on the
scientific function of THMC coupling in the studies of CO2 injection induced seismicity, so as to provide reference
and guidance for the researches on the mechanism analysis and forecasting of such induced seismicity.

greenhouse gases, carbon capture, utilization and storage (CCUS) tech­


1. Introduction nology can directly achieve the goal of carbon reduction (Aminu et al.,
2017). Relevant theoretical researches and technical methods need to be
Climate change has become a worldwide common challenge. In conducted urgently. CO2 geological storage (CGS) is an important part of
recent decades, the emissions of greenhouse gas result in temperature CCUS. The CO2 produced by fixed point sources (mostly from industries
increase, sea level rise, frequent extreme weather and climate events such as power plants) is firstly collected and liquefied, then the CO2 is
(Geneva, 2013). Reducing greenhouse gas emissions represented by CO2 injected into closed deep geological structures for permanent storage to
has become a consensus all over the world (Eggleston et al., 2006; reduce the man-made CO2 into atmosphere. This process requires a large
Nordhaus, 1992). Many countries also actively take measures to deal amount of high-pressure CO2 fluid to be injected underground, which
with this problem (Schneider, 1989). As an effective means to reduce will affect the original mechanical equilibrium state of the reservoir,

* Corresponding author at: Key Lab of Groundwater Resource and Environment, Ministry of Education, Jilin University, Changchun, China.
E-mail address: fengbo82@126.com (B. Feng).

https://doi.org/10.1016/j.earscirev.2023.104369
Received 3 December 2022; Received in revised form 2 February 2023; Accepted 27 February 2023
Available online 11 March 2023
0012-8252/© 2023 Elsevier B.V. All rights reserved.
Y. Cheng et al. Earth-Science Reviews 239 (2023) 104369

Nomenclature S Storage coefficient


Rt Induced seismicity trigger front
D Hydraulic diffusivity τ0 Stressing rate
τs Shear strength ta Characteristic decay time
σn Normal stress Ds Seismic hydraulic diffusivity
f Friction coefficient K Bulk modulus
P(t) Pore pressure perturbation αT Linear thermal expansion coefficient
φ Friction angle σ 1, σ2, σ3 Principal effective stresses
c Cohesion Pa Mean principal effective stress
σ1 Maximum principal effective stress fp Fracture potential
σ3 Minimum principal effective stress PC Critical differential stress
C Critical pressure G Modulus of rigidity
Cmax Maximum critical pressure ΔV Injection volume
|r| Distance from seismogenic point to injection point Σ Seismogenic index
q Injection flow rate NM Seismic quantity
t Injection time M Magnitude
erfc Complementary error function Mc Reference magnitude
Nev Seismic event quantity b Gutenberg-Richter exponent
Pi Injection pressure Q Cumulative injected volume
Nb Bulk concentration
Rt The radius of the triggering front Abbreviations
Sij, Sii, Sjj Initial pre-stress field CCUS Carbon capture, utilization and storage
sij, sii, sjj Incremental stresses CGS CO2 geological storage
ρ Rock density EOR Enhanced oil recovery
u, v Incremental displacements CO2-EOR CO2-enhanced oil recovery
γ Seismicity Enhancement Factor EGS Enhanced Geothermal System
CRST Lower pore pressure threshold CPT Critical pressure theory
τtec Tectonic shear stress rate THMC coupling Thermo-hydro-mechanical-chemical coupling
α Dimensionless state parameter FP Fracture potential
vtec tectonic background seismicity PSHA Probabilistic seismic hazard assessment
k Permeability IoT Internet of things
μ Viscosity

cause perturbation, trigger seismicity and form negative environmental However, there is still a huge knowledge gap in the research of CO2
risks (Nicol et al., 2011). injection induced seismicity, including how engineering parameters
So far, many CO2 injection projects have monitored induced seis­ (injection flow rate, pressure, time, volume etc.) and reservoir proper­
micity (White and Foxall, 2016). And the CGS usually requires high ties (lithology, fault properties, in-situ stress levels etc.) affect induced
porosity and permeability reservoirs for CO2 perfusion and longtime seismicity, reasonable fault activation simulation method and the cor­
storage. This kind of reservoir is often convenient for the flowing of CO2 relation analysis about large earthquakes, seismic risk mitigation
and pressure diffusion. Compared with tight reservoir, CO2 plume method, magnitude upper limit key control factors. The lack of solutions
diffusion has less perturbation to the original reservoir stresses, and the for these problems have impeded the CGS development. Therefore, this
triggered induced earthquake magnitudes are small, generally below review focuses on these research challenges by expounding recent
one magnitude (Vilarrasa et al., 2019). Although it can be monitored, it outstanding issues and major scientific advances, in order to provide an
is usually not enough to be felt by humans. Hence, the CO2 injection comprehensive overview of scientific progress about the CGS induced
induced seismicity has been known for decades, the risk is downplayed seismicity.
in early researches (Damen et al., 2006). However, in recent years, the
induced seismicity caused by wastewater injection, shale gas hydraulic 2. Field experience of CO2 injection induced seismicity
fracturing and Enhanced Geothermal System (EGS) development has
attracted more attention, and some projects have even trigger large Compared with wastewater injection, shale gas hydraulic fracturing
magnitude seismic events. Since the injected CO2 volume of the pro­ and EGS, the CGS usually operates in high permeability reservoirs for
posed commercial scale CO2 storage projects always significantly ex­ CO2 long time migration and storage, the injection pressure is easier to
ceeds the injected fluid volume in such cases, and the seismic risk dissipate, the increase of injection pressure is always smaller under the
induced by CO2 injection needs to be reassessed (White and Foxall, same injection volume. Therefore, the magnitudes of the induced seis­
2016; Zoback and Gorelick, 2012). micity in CGS are generally smaller than those in other injection oper­
For CGS project, the induced seismicity risks include: i) the local ations. On the other hand, the CO2 in CGS migrates farther due to high
community will pay attention to the perceptible induced seismicity, and reservoir permeability, the spatial distribution of the induced seismicity
large earthquakes will cause casualties; ii) the infrastructure may be in CGS is usually wider. Although most seismic events related to CO2
damaged, especially when storage site is close to urban areas; iii) injection are microseismic events (magnitude M < 3), large magnitude
induced seismicity may destroy the caprock and CO2 will migrate to the events are occasionally recorded. Since 2004 the CO2-enhanced oil re­
ground surface along the damaged zone. To mitigate the induced seis­ covery (CO2-EOR) project in the Cogdell, United States injected 85
micity risks of CGS projects, it is necessary to identify the triggering million m3 gas per month into deep layer, eighteen M ≥ 3 earthquakes
mechanism, maximum magnitude determinants, temporal and spatial were recorded between 2006 and 2011, of which the maximum
distribution characteristics and magnitude prediction methods. magnitude event was an M4.4 earthquake in 2011 (Davis and

2
Y. Cheng et al. Earth-Science Reviews 239 (2023) 104369

Pennington, 1989; Gan and Frohlich, 2013b). The CGS project in Dec­ shown in Table 1.
atur, Illinois basin injected supercritical CO2 at the rate of 100 million Aneth: As Utah’s largest oil producer, Aneth oil field is selected to
tons per year, resulting in the increase of reservoir pore pressure over a demonstrate that the combination of the enhanced oil recovery (EOR)
region of 40,000 km2, and the stress perturbation induced by the CO2 and CO2 geological storage is feasible. The induced seismicity was
plume might influence the Wabash fault zone, which might relate to the monitored during CO2-EOR operation from late 2007 to July 2009 with
M5.2 natural earthquake in 2008 (Zoback and Gorelick, 2012). Zoback a downhole array of geophones (Soma and Rutledge, 2013). About 3800
and Gorelick (2012) and Rutqvist et al. (2016) warn that large-scale CO2 events with magnitudes between − 1.2 to 0.8 were recorded within
injection could cause a wide range of rock mass stress perturbation, about 4.8 km, and these seismic events distributed along two faults
trigger tectonic fault activation, and have a high possibility to induce (White and Foxall, 2016). On June 6, 2008, an M3.7 earthquake
large earthquakes. The American National Research Council also pro­ occurred about 9.5 km away from Aneth Unit (Rutledge, 2011). Even
posed that large volume fluid injection into deep underground for a long though this large magnitude event was not directly related to CO2 in­
time has great potential to induce large magnitude seismic events jection, the stress perturbation caused by CO2 plume might promote the
(Council, 2013). The CGS projects often involve large-scale and long- triggering of large earthquakes.
term supercritical CO2 injection. Then, the induced seismicity of CO2 Cogdell: Between 2004 and 2011, large amount of CO2 was injected
injection receives attention in recent years. In addition, even small to underground to enhanced oil production at Cogdell oil field north in
moderate magnitude earthquakes may threaten the integrity of caprock, Texas, USA. Month injected gas volumes exceeded 85 million m3. The
resulting in the risk of CO2 leakage (Mazzoldi et al., 2012). Therefore, it injected gas was a mixture of CO2 and methane which caused a series of
is important to understand the field experience of CO2 injection projects. earthquakes. A total of 105 events with magnitude M ≥ 1 were recorded
The induced seismicity related to on-site CO2 injection experience is by regional seismic network during CO2-EOR operation between 2009
and 2011 (Gan and Frohlich, 2013a). It is the only officially reported
Table 1 earthquakes induced by CO2 injection which have the magnitudes larger
Summary of observed induced seismicity at CO2 injection projects. than M1 (Gan and Frohlich, 2013b). The permanent regional network
recorded 8 seismic events have magnitudes between M3 and M4.4
Project Location Category Observed Other important
magnitude information (White and Foxall, 2016). In addition, a M4.6 event was also recorded
during water injection between 1956 and 1982 for EOR at Cogdell field
Aneth USA CO2-EOR M-1.2 to Recorded seismic
M0.8 event quantity: (Davis and Pennington, 1989).
about 3800 Weyburn: The Weyburn-Midale CO2-EOR project is the world’s
events. earliest seismicity monitoring and largest CGS project (Aminu et al.,
An M3.7 seismic 2017). The injected CO2 reached about 5.3 million tons per year
event may relate
(Whittaker et al., 2011). A deep downhole seismic monitoring system
to the CO2
injection. was installed in an abandoned production well to record induced seis­
Two faults are micity activity (Verdon et al., 2010). About 200 seismic events were
found in the detected with the magnitude between M-3 and M-1 over seven years, the
reservoir.
seismic frequency and magnitude are extremely low (White and Foxall,
Cogdell USA CO2-EOR One M4.4 Recorded seismic
event, event quantity:
2016). Simulation results shows that these micro-seismic events are
eighteen M ≥ 105 seismic events induced by in situ stress perturbation caused by reservoir deformation,
3 events with magnitude M rather than CO2 injection induced pressure perturbation (Verdon et al.,
≥ 1. 2011).
So far the recorded
Decatur: The Decatur project in the Illinois basin is the first large-
largest magnitude
seismic event scale dedicated CGS project in America. From 2011 to, 2014, approxi­
induced by CO2 mately one million ton CO2 was injected into a 550 m thick, highly
injection permeable sandstone reservoir at 2100 m depth (Couëslan et al., 2014).
Weyburn Saskatchewan, CO2-EOR, M-3 to M-1. Recorded seismic
Two induced seismicity monitoring arrays were installed in two deep
Canada Oil and gas event quantity:
reservoir about 200 events.
wells (Bauer et al., 2016). Another monitoring system includes 13
Decatur Illinois, USA CGS M-2 to M1 Recorded seismic broadband seismic stations at the surface and an array with 4 acceler­
event quantity: ometers in shallow borehole (Kaven et al., 2015). More than ten thou­
>10,000 events. sand seismic events with magnitudes between M-2 and M1 were
Multiple faults.
recorded in 3 years. The seismic analysis results showed that some small
In Salah Tamanrasset, CGS, Oil M-1 to M1.7. Recorded seismic
Algeria and gas event quantity: faults were reactivated due to the stress perturbation caused by CO2
reservoir >6000 events. plume (White and Foxall, 2016). With the increase of injection time, the
Injection well spatial distribution range of induced seismicity also increased gradually.
passes through a In Salah: The Algerian In Salah CGS project is jointly operated by a
fault, and CO2
injection triggers
consortium from the Algeria, UK and Norway. It is estimated that the
multiple minor CO2 storage capacity of the In Salah oil field Carboniferous Krechba
faults. sandstone reservoir is about 17 million tons (Mathieson et al., 2011).
Lacq- France CGS, Oil M-2.3 to M- Recorded seismic From 2004 to 2011, about 4 million tons of CO2 was injected under­
Rousse and gas 0.5. event quantity:
ground for storage (Aminu et al., 2017). The sandstone reservoir is about
reservoir about 2500
events. 20 m thick with the porosity of 13%–20%, averaging permeability of 10
Only over 600 mD, and has an approximately 950 m thick mudstone caprock. This low
events are located. porosity and permeability will easily lead to the increasing of pore
Otway Victoria, Oil and gas M<0 Recorded seismic pressure and geomechanically deformation, which has already been
Basin Australia reservoir event quantity: <5
events per day.
confirmed by InSAR monitoring (Onuma and Ohkawa, 2009). The
CO2 and CH4 induced seismicity monitoring arrays were installed in three injection
mixed wells, >6000 seismic events with the magnitude from M-1 to M1.7 were
supercritical fluid detected (Verdon and Stork, 2016). The 3D seismic analysis shows that
were injected
the injection well passes through a fault, and CO2 injection triggers the

3
Y. Cheng et al. Earth-Science Reviews 239 (2023) 104369

activation of minor faults, resulting in the seismic spatial and temporal that different seismic patterns may occur with similar geological profiles
distribution along the faults (Rutqvist et al., 2016). and injection conditions. In this poor data circumstances, it will be
Lacq-Rousse: Lacq-Rousse CGS project is locate in Southwest of necessary to conduct seismic monitoring system in future CO2 storage
France. In three years, >50 thousand tons of CO2 were injected into a projects. The increasing field data will help researchers identify and
4500 m deep dolomitic reservoir. A hybrid seismic monitoring network manage seismicity risks before induced seismicity become a major
was installed, which contained a near-surface seismic monitoring array, obstacle to the development of the CGS.
and a downhole monitoring array in an injection well. About 2500
seismic events were recorded over 3 years. Due to the lack of enough 3. Triggering mechanism of CO2 injection induced seismicity
energy, the location and magnitude of most seismic events cannot be
determined. Only over 600 events, with the magnitudes between M-2.3 The seismicity induced by CO2 injection includes three scenarios
and M-0.5 were located (Payre et al., 2014). (Fig. 1): i) no large-scale tectonic fault is in the reservoir (Fig. 1 a),
Otway Basin: The Otway Basin CGS project is locate in approxi­ pressure buildup caused by CO2 plume breaks the rock around wellbore
mately 300 km south-west of Victoria, Australia (Dance, 2013). A total and triggers induced seismicity; ii) The CO2 injection point has a certain
of 65,445 tons CO2 and CH4 mixed supercritical fluid were injected into distance from the fault (Fig. 1 b), the CO2 plume or stress perturbation
a 2000 m deep reservoir (Hortle et al., 2011). Preliminary seismic risk reach the fault, the fault may be activated and the induced seismicity is
analysis results show that the induced seismicity activity related to CO2 triggered near the injection point and on the fault; iii) CO2 is directly
injection should be very low, and no large magnitude event will be injected into the fault (Fig. 1 c), the fault is activated and earthquakes
trigger (Stirling et al., 2011). On site seismic monitoring also reveals that are triggered. For the case 1, without fault, all induced earthquakes are
a low level of induced seismicity, the magnitude is usually <0, the caused by rock fracturing during injection. The reservoirs of CGS pro­
number of seismic events per day is usually <5 (Daley, 2009; Myer and jects are generally sedimentary rocks, which are much more porous and
Daley, 2011). softer than crystalline rocks. The injection perturbation pressure is
Sleipner: The Norwegian Sleipner project is the first commercial easier to dissipate in this kind spongy sedimentary reservoir. Therefore,
CGS in the word, which is locate in North Sea (Torp and Gale, 2004). in most circumstance, the CO2 injection only causes micro-seismicity
>15.5 million tons CO2 had been injected into the saline aquifer which is with a low magnitude (M < 2) which could not be felt on the ground
approximately 1000 m below the sea floor (Aminu et al., 2017). The surface, some CO2 injection projects event could not detect induced
induced seismicity monitoring method is not conducted in this CO2 in­ seismic events. For the case 2, during CO2 injection, pressure pertur­
jection operation. On the other hand, the storage reservoir is a sandstone bation firstly causes rock damage around the wellbore and only micro-
layer with high permeability. Hence, no induced seismicity is record seismicity is triggered in this stage. Then, supercritical CO2 fluid grad­
(Verdon and Stork, 2016). ually diffuses to the fault, when the fault is activated, large magnitude
Snøhvit: The Snøhvit CGS project is also locate in Norway, operated events could be triggered. For the case 3, the fault activation trends to
by Statoil (Aminu et al., 2017). Unlike reservoir of Sleipner, the trigger large magnitude events, which could be detected by the seismic
permeability of sandstone reservoir at Snøhvit is low, and the pore monitoring system, the catastrophic M > 3 events may even be
pressure increases significantly during CO2 injecting (Chiaramonte triggered.
et al., 2015). Since no induced seismicity monitoring method is used, no The triggering mechanism of induced seismicity is that the CO2 in­
seismic event is recorded, either (Verdon and Stork, 2016). jection leads to pressure buildup, reduces the effective stress, destroys
Ketzin: The Ketzin CGS project is locate 40 km west of Berlin, Ger­ the original region stress equilibrium state in the reservoir, causes rock
many (Le Guen et al., 2011). The reservoir is an 80 m thick saline formation mechanical failure, and releases the energy stored in the rock
sandstone aquifer with a depth of 630 m. About 70 thousand tons CO2 mass in the form of seismic waves (Vilarrasa et al., 2019). In recent
were injected from 2008 to 2009 (Aminu et al., 2017). After the years, several theories of induced seismicity mechanism have been
completion of operation, the pressure at injection depth increased by proposed, the ones could be used for CO2 injection are list in Table 2.
about 1.7 MPa (Würdemann et al., 2010). No seismic events from Ketzin The first theory is called critical pressure theory (CPT) which is
site were detected by the regional seismic network (Evans et al., 2012). proposed by Shapiro (2015). In this theory, reservoir is porous media,
Cranfield: The Cranfield CO2 storage project is carried out on the
site of Cranfield oilfield which is in Mississippi, USA. The reservoir is
saline Tuscaloosa sandstone (MIT, 2022). About 5 million tons of CO2
were injected into the target saline aquifer (Freifeld et al., 2013). A
seismic monitoring system was installed in shallow wells, no induced
seismicity was detected (Hovorka et al., 2013).
Aquistore: The Aquistore CGS site is locate in the intracratonic basin
of Northern Saskatchewan, Canada. The reservoir is Cambro-Ordovician
sandstone and the basement is Precambrian granite. The natural seis­
micity in this region is low. Since April 2015, about 140 thousand tons
CO2 was injected into a 3.2 km deep underground reservoir, and an
extensive earthquake monitoring network was set up. It includes a small
passive broadband seismic monitoring network, a near surface contin­
uous recording seismograph array, and a temporary downhole geophone
in the depth about 3000 m. The monitoring results show that the pore
pressure perturbation is very small, no significant induced seismicity is
detected (Stork et al., 2018).
Although the above field experiences have improved our under­
standing of specific site induced seismic hazards, the overall field
Fig. 1. The possible induced seismicity scenarios of CO2 injection: (a) no fault,
seismic database of CO2 injection is very small. Due to the limited data,
the seismicity is caused by the rock fracturing process due to injection; (b) CO2
it is difficult to draw any deterministic conclusions about the seismic injection is away from the fault, the pore pressure migrates to the fault, both
risks associated with CO2 injection. In addition, the seismic triggering rock fracturing and fault activation could trigger induced earthquakes; (c)
mechanics is highly dependent on reservoir geological characteristics injecting CO2 into fault, this is similar to the scenario (b) (modified from Els­
and injection parameters, even the field cases described here also show worth et al., 2016).

4
Y. Cheng et al. Earth-Science Reviews 239 (2023) 104369

Table 2 Table 2 (continued )


The main triggering mechanism theories related to injection induced seismicity. Theory Summary
Theory Summary
where is Pa the actual differential stress, PC is the critical
Critical pressure 1) The potential seismic events are determined by randomly differential stress.
theory (Rothert distributed pre-existing defects in the reservoir which are 2) The actual differential stress is the mean principal effective
and Shapiro, called critical pressures (C). stress:
2007; Shapiro, 2) For a specific point |r| in the reservoir, the pressure Pa = (σ1 + σ2 + σ3)/3
2015) perturbation where σ1, σ2, σ3 are the principal effective stresses.
( )
q |r| 3) The critical differential stress is defined as:
p(t)∝ erfc √̅̅̅̅̅ PC = 2(c cos φ + Pa sin φ)
D|r| Dt
where |r| is the distance between injection point and a where c is the cohesion; φ is the angle of friction.
specific point, q is the injection flow rate, D is hydraulic The fracture potential has a positive correlation with shear
diffusivity, t is the injection time, erfc(x) is the failure and induced seismic activity.
complementary error function.
3) Seismic event quantity (Nev) highly depends on the
injection pressure perturbation, and is determined by and contains randomly distributed pre-existing defects (or small faults).
Gutenberg-Richter magnitude frequency distribution: Each defect is a seismogenic element, and is given for an individual
DNb Pi t
Nev ∝ critical pressure to compare with the pore pressure perturbation. Only
Cmax
where Pi is the injection pressure, Nb is the defect bulk one seismic event could be triggered in a defect. It means that the
concentration, Cmax is the maximum critical pressure. density of defects (defects per unit volume) determines the density of
4) The seismicity triggering front (the distance between the seismicity. The distribution of the defects in the reservoir is often
farthest induced seismic event and fluid injection point):
√̅̅̅̅̅ assumed to follow a normal distribution. It is assumed that induced
Rt ∝ Dt
It means that the radius of the triggering front (Rt) is
seismicity is triggered by pore-pressure perturbation. An earthquake
determined by hydraulic diffusivity. happens when pore pressure exceeds the critical values. Then the spatial
Biot’s incremental When neglecting body forces, the displacement due to rock distribution rang of seismic events is directly related to the range of
strain theory ( formation deformation could considered as plane-strain pressure perturbation. The Coulomb failure criterion is used to deter­
Kisslinger and problem:
mine whether a defect or small fault will transform into a seismic event
Cherry, 1970; ∂sii ∂sij ∂w ( ) ∂w ∂Sii ∂Sij
Kisslinger, 1976)
ρü = +
∂x ∂y
− 2Sij + Sii − Sjj
∂x

∂y ∂ x
exx −
∂y
eyy − (Shapiro et al., 2007). Similarly to the tectonic seismicity, the induced
(∂S
ii ∂Sij
) seismicity defined by CPT is also compatible with the Gutenberg Richter
e
magnitude frequency distribution (Shapiro, 2015). In CPT method, the
+
∂y ∂x xy
∂sij ∂sjj ∂w ( ) ∂w ∂Sjj ∂Sij induced seismicity is positively correlated with the injection pressuri­
ρv̈ = + + 2Sij + Sii − Sjj − e − e −
∂x ∂y ∂y ∂x ∂y yy ∂xx xx
( ∂S ) zation, this method postulates induced seismicity is not related to tec­
jj ∂Sij
∂x
+
∂y
exy tonic background. In field, this method can well estimate and explain the
where Sij, Sii, Sjj, are the initial pre-stress field; sij, sii, sjj, are the quantity and temporal-spatial distribution of induced seismicity. When
incremental stresses induced by fluid injection; ρ is the rock simulating seismic activity, there is no time delay between the pressure
formation density; u, v, are the incremental displacements change and the triggering of seismicity (Wenzel, 2017). However, the
caused by reservoir deformation, these two displacement
CPT theory only considers the change of normal stress, and postulates
components express the shear strain variation induced by
injection behavior in rock formation. induced seismicity is not related with tectonic background. Due to the
Rate- and state- 1) The Seismicity Enhancement Factor γ(t) defines the lack of consideration of large-scale tectonic faults, this method is
dependent increase of induced seismicity compared to the natural impossible to predict the perceivable earthquakes induced by the shear
frictional theory ( tectonic seismicity under the influence of injection pressure
failure of pre-existing faults. This is also the limitation of CPT method.
Dieterich, 1994; perturbation, it is the state variable for induced seismicity.
Wenzel, 2015) dγ 1
(
p(t)
( (
p(t)
) The reservoirs of CGS projects are usually porous sandstone formation,
= • 1+ − γ(t) • τ̇tec • 1 + + g0 • and if the existence of large tectonic faults could be avoided in the early
dt CRST σn σn
exploration stage, most of the induced seismicity triggered by CO2
))
ṗ(t)
plume would be small and imperceptible, then the CPT method will be
where CRST is the lower threshold of pore pressure
suitable for the estimation and analysis of induced seismicity in these
perturbation p(t), below which no seismicity is induced. It is
similar to the critical pressures in CPT theory; σn is the normal situations. Verdon et al. (2015) used this method to establish a geo­
stress, τtec is tectonic shear stress rate; g0 = (τtec/σn − α), α is a mechanical model to simulate the CO2 injection induced seismicity
dimensionless state parameter, which represents the monitored in the In Salah CGS project, Algeria. The pre-existing frac­
dependence on the evolution of fault normal stress and could tures were assumed as the defects in CPT method, the seismicity spatial-
be got from laboratory experiments (Linker and Dieterich,
1992).
temporal distributions and seismic triggering were simulated based on
2) The seismic density (Ṅev ) that exceeds a certain magnitude the fracture locations, orientations, sizes and pressure perturbation. The
can be expressed as: simulation results of seismic locations, activity rates and magnitudes are
Ṅev (t) =
1 v̇tec

in good agreement with the observations, and the predicted largest
γ(t) τ̇tec magnitude event also well matched the observed records.
where vtec is the tectonic background seismicity per unit
reservoir volume.
The Biot’s incremental strain theory was proposed by Kisslinger and
3) Then the total seismicity induced by pore pressure in the Cherry (1970), which could analyze fault activation and seismicity
whole reservoir volume:

triggering mechanism induced by the shear stress variation associated
Nev (t) = ṅ(t)dV with fluid injection. This theory holds that when the seismicity is
Remark: The induced seismicity in this theory is got based on induced by rock formation breaking, the regional tectonic stress should
the tectonic background seismicity, reservoir stress rate and
state evolution, and pore pressure perturbation.
be close to the rock breaking strength before injection. When the in­
Fracture potential 1) The fracture potential (fp) describes the relationship jection pressure perturbation plus the tectonic stress exceeds the rock
theory (Verdon between the stress state and the Mohr Coulomb failure curve. strength, the rock breaks and induced seismicity is triggered. Since the
et al., 2011) In fault activation, it can be defined as the ratio of the actual induced seismicity is triggered due to the activation of pre-existing
differential stress to the failure critical differential stress:
tectonic fault, the fault should be subjected to substantial shear stress
Pa
fp = before fluid injection, and the shear stress is still lower than the shear
PC
strength of the fault, the shear stress is not enough to activate the fault.

5
Y. Cheng et al. Earth-Science Reviews 239 (2023) 104369

This means that the fault is in a critical stress state, and a small Similar to CPT theory, this method also has a critical slip size, when the
perturbation may trigger fault activation. Injection behavior may in­ size is exceeded, the fault will become unstable and the slip rate is very
crease the shear stress or reduce the shear strength of the fault surface. In high. This theory requires a specific tectonic background to analyze the
addition, the increase of pore pressure due to injection may also reduce slip rate and the response of stress field state, the seismic activity in this
the effective normal stress. All these three circumstances could drive the theory obeys the Omori’s law, and the general seismicity equation is
slip of fault. The Biot’s theory explains the effect of pre-stress variations derived based on the tectonic background seismicity, reservoir stress
on shear strain and deformation, and depicts the shear stress concen­ change rate, stress state evolution history background, and the effect of
tration caused by shear wave. This can produce tensile stress which will pore pressure perturbation. Therefore, this theory could be used to
reduce normal stress and weaken the fault. At the same time, the analyzed injection induced seismicity without Coulomb failure crite­
reservoir permeability should be within a reasonable range. The rion. An earthquake simulator called RSQSim is developed based on the
permeability should be high enough to inject fluid, and low enough to Dieterich theory (Richards-Dinger and Dieterich, 2012), which could
make pore pressure buildup possible. The sandstones, most CGS reser­ simulate the seismic time-dependent clustering phenomena and seismic
voirs, meet this requirement. In this theory, the seismicity triggering temporal sequences according to the reservoir physical properties and
mechanism satisfies the effective stress theory and Mohr-Coulomb shear tectonic seismicity statistical characteristics (Dieterich et al., 2015), and
failure criterion (Kisslinger and Cherry, 1970). The CGS reservoir is therefore this simulator is effective for the seismicity induced by CO2
usually a porous medium, CO2 injection will increase the pore pressure injection (Kroll et al., 2020). Kroll et al. (2020) established numerical
and reduce the effective stress, the Mohr circle will move towards left. models with RSQSim to analyze the change of effective normal stress
When the circle is tangential to the failure envelope, the fault is acti­ when CO2 was injected into saline aquifer, simulate the seismic response
vated and shear failure occurs (Fig. 2). The long-term injection of CO2 under the different tectonic fault scenarios, and assessed the risk of
will cause chemical reaction between CO2 and minerals and weaken the induced seismicity under different injection schemes.
strength of faults or rocks. However, this method completely neglects Recently, a new method called fracture potential (FP) theory was
the failure mechanism caused by this stress corrosion process. proposed to analyze the seismicity induced by CO2 injection (Verdon
The Dieterich rate- and state-dependent frictional theory assumes et al., 2011). The FP was initially used for the evaluation of fracture
that induced seismicity is proportional to the tectonic background permeability and fluid flow in experiments (Connolly and Cosgrove,
seismicity (Dieterich, 1994). Compared with the CPT method, the 1999). The key calculation steps of FP method are shown in Table 2. In
Dieterich theory is more complex, and both the changes of normal and this theory, the FP describes the approach of the stress state to the Mohr
shear stress are considered. And the same as the Biot’s incremental strain Coulomb curve, thus expressing the possibility of shear failure and
theory, the Dieterich theory also considers that the change of friction on evaluating the induced seismic activity. The injection and flow of CO2 in
the fault surface leads to the fault slip, thus triggering the induced the reservoir are the important reasons of FP changes and anomalies.
seismicity. In this theory, more presuppositions are also needed, the Verdon et al. (2011) simulated the FPs during CO2 injection at Weyburn,
fault always keeps slip without rest, when fluid is injected, pore pressure Canada; the numerical model was established by combining a fluid flow
increases, then the geologic slow aseismic slip may transform to seismic simulation software called MORE (by Roxar Ltd) with a geomechanical
slip; at the same time, the seismogenic fault needs to slip under a con­ software called ELFEN (Rockfield Ltd) through a Message Passing
stant tectonic shear stress condition, the initial slip rate on the fault Interface (MPI, also by Rockfield Ltd). The results show that the
surface should has a non-uniform distribution; these presuppositions anomalies of FP could localize the migration of pore pressure, thus
will create a time-dependent constant tectonic seismicity background effectively monitoring the CO2 plume expansion area, flow path and
(Dieterich, 1994). The slip process from aseismic to seismic determines evaluating the seismic risk. Fig. 3 shows the simulation results of FPs in
the evolution of state parameters, the state parameters and slip rate the reservoir and caprock. The FP anomalies tightly correlates with the
determine the evolution of friction on the fault surface (Wenzel, 2017). CO2 injection and migration.
Although these mechanisms of injection induced seismicity have
been proposed in recent years, these theories rely on assumptions and
input parameters. Sometimes, these prerequisites may be different from
the field conditions, making it difficult to accurately estimate the on-site
seismic activity.

4. Fault activation analysis induced by CO2 injection

The fault activation and induced seismicity are inevitable and


important issues for CGS from security and public acceptance perspec­
tives. The injection behavior will cause the pressure accumulation in the
reservoir and change the natural in-situ stress state, once the fault is
activated or rock is damaged, the stored energy in the formation due to
long-term accumulation of crustal stress will be released in the form of
seismic waves. Many recent studies about CGS projects showed that
large-scale CO2 injection will promote fault reactivation, such as the
Ohio River Valley deep saline reservoir CGS project in West Virginia, US
(Lucier et al., 2006), the Teapot Dome CO2-EOR project in Wyoming, US
(Chiaramonte et al., 2008), the Saint-Martin de Bossenay oil field CO2-
Fig. 2. The shear failure process showed by the Mohr circle, the CO2 injection EOR project in Paris Basin, France (Vidal-Gilbert et al., 2009), a CO2-
increases pore pressure, reduces the effective stress, then the Mohr circle moves
EOR project in Brazil (Mendes et al., 2010), the In Salah CGS project in
to the left, shear failure happens when the circle reaches the point (red) of
Algeria (Rutqvist et al., 2010), the Po River CO2 sequestration project in
tangency with the failure envelope. τs is the shear strength, σ n is the initial
effective normal stress, Pinj is the pore pressure required for shear failure, φ is gas fields, Italy (Ferronato et al., 2010), a CO2 sequestration project in an
the friction angle, c is the cohesion, σ1 and σ3 are the maximum and minimum offshore gas field in Netherlands (Orlic et al., 2011), the Snøhvit CGS
principal effective stresses, respectively. (For interpretation of the references to project in Norway (Chiaramonte et al., 2011); the Otway CGS project in
colour in this figure legend, the reader is referred to the web version of Australia (Vidal-Gilbert et al., 2010), and the Ketzin CGS field experi­
this article.) ment in Germany (Ouellet et al., 2011). In addition, these studies also

6
Y. Cheng et al. Earth-Science Reviews 239 (2023) 104369

Fig. 3. The simulation results of FP anomalies


induced by CO2 injection at Weyburn, the vertical
injection wells are marked by triangles, the horizontal
producing wells by black lines, high FP represents
that the seismic activity should also be high: (a) is the
FP percentage changes versus time, the solid lines
represent the FP percentage changes in the reservoir,
the dashed lines represent the FP percentage changes
in the caprock. FP increases in the caprock near the
injection wells (timestep 11, black dashed line),
therefore this region microseismicity is more active.
(b) is the FPs in the reservoir after one year of CO2
injection, the PFs near the production wells are higher
than those near the injection wells, majority of indued
earthquakes occur close to production wells in the
reservoir. (c) is the FPs in the caprock after one year
of CO2 injection, the FPs near the injection wells are
higher, the seismicity is more active in this region.
The simulation results well match the observed
microseismicity (modified after Verdon et al., 2011).

show that the fault activation induced by CO2 injection always concerns
a complex thermo-hydro-mechanical-chemical (THMC) coupling pro­
cess, which include pore pressure perturbation, temperature alteration
and stress corrosion. The CO2 injection and its non-uniform flowing
diffusion process in the reservoir may cause pore pressure to perturb the
in-situ stress field, this results in the hydro-mechanical-coupling effect.
The temperature of supercritical CO2 is far lower than the original
temperature of the reservoir, the sudden injection of CO2 will reduce the
surrounding rock temperature, induce thermal stresses and thermal
fractures, change the in-situ stress field and affect fault stability. This is
the thermo-mechanical coupling. The long term injection of CO2 will
change the chemical environmental in the reservoir, the CO2 will react
with minerals, thus changing rock strength and fault friction coefficient,
this long term chemical action will change the in-situ stress state in the
formation, which may form fractures and lead to fault activation. This is
the chemical-mechanical coupling. Generally speaking, during CO2 in­
jection, each coupling process will not occur alone, but affect each other.
Therefore, in the THMC coupling analysis, we need to ascertain the ef­
fects of multi-parameter coupling evolution on the fault stability, esti­
mate the impact of normal stress and shear stress evolution on the fault
stress state, and evaluate and characterize fault activation process and Fig. 4. The diagram fault activation associated with CO2 plume perturbation,
induced seismicity activity. This is also the fundamental knowledge for the blank arrows indicate the direction of formation deformation, and the gray
region indicates the pressurized areas caused by the pore pressure accumula­
fault activation analysis.
tion, dark arrows represent the compression direction.

4.1. Pore pressure perturbation hydraulic characteristics of the formation, that is, the hydraulic diffu­
sivity (D).
Pore pressure perturbation is the main triggering mechanism of CO2 Pore pressure perturbation changes the effective stress state and
injection induced seismicity, it includes pore pressure accumulation and leads to fault activation. The triggering of large magnitude events are
diffusion. Once the CO2 is injected into the reservoir, the pore pressure generally caused by the tectonic fault activation. The seismic triggering
accumulate near the fault, when the CO2 plume migrate to the fault, the mechanism due to fault activation is mainly controlled by two physical
pore pressure will diffuse along the fault, the friction resistance on the factors: the pore pressure diffusivity and the effective stress state on the
fault surface will decrease in the influence of pore pressure perturbation. fault. Therefore, the key to analyze fault activation is to determine the
If the fault is in a critical state, a small increase in pore pressure can pore pressure perturbation along the fault. And Shapiro et al. (2003)
cause fault activation by reducing the effective normal stress. Fig. 4 described by this pressure perturbation P(t, r) process based on the hy­
schematically describes the perturbation of CO2 plume in the reservoir. draulic diffusivity (D), which was defined by the following equation:
Due to the accumulation of pore pressure from the injection zone, the
reservoir is compressed in the pressurized region (gray). The increase of ∂P(t, r)
= D∇2 P(t, r) (1)
pore pressure in the reservoir leads to reservoir expansion. This expan­ ∂t
sion leads to the horizontal extension of the basement and caprock rocks where r is the radius vector from the injection point to the observation
with the increase of pore pressure. At the same time, the lateral point, t is the time. This equation can express the pore pressure diffusion
expansion of the reservoir leads to the horizontal compression of the with time in porous media, give the perturbation pressure at a point in
basement and caprock at the region far away from the injection zone. the reservoir, and has been used in many injection induced seismicity
Therefore, even if the pore pressure does not reach, this extension/ works. The hydraulic diffusivity D = k/μS, where is the k formation
compression in the formation perturbs the stress state of the faults. The permeability, μ is the fluid viscosity and S is the storage coefficient.
boundary of the extensional/compressional region is determined by the On the basis of the research of Shapiro et al., Chang and Segall
pore pressure diffusion process and perturbation apogee, and thus by the

7
Y. Cheng et al. Earth-Science Reviews 239 (2023) 104369

(2016) analyzed the general process of this fault activation induced by hydraulic diffusivity:
CO2 plume perturbation. As shown in Fig. 5, the pore pressure diffusion √̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
in the pressurized region is determined by the hydraulic diffusivity of M = 1− erf (Ds /4D) (4)
rock formation. Once the pressurized CO2 plume approaches or reaches
where Ds is a parameter called “seismic hydraulic diffusivity”, which
the fault, the plume will cause the local pressure gradient and distort the
define the induced seismicity distributions controlled by pore pressure
pore pressure contours near the fault. The change of stress state is mainly
diffusion. erf is the error function. It can be seen that the hydraulic
controlled by the plume into the fault, and the fault itself acts as a virtual
diffusivity (D) is a key parameter in the estimation of injection induced
linear pressure diffusion source for the plume to spread into the sur­
seismicity. Fig. 6 and Eqs. (1)–(4) shows the application of hydraulic
rounding strata. The plume flowing into the fault will also attenuate the
pore pressure accumulation in reservoir. Besides, it is noteworthy that
the pore pressure perturbation also depends on fault morphology
(location, length, dip and aperture, etc.), in-situ stresses, injection op­
erations (pressure, duration and rate, etc.) and formation porous elastic
parameters.
Shapiro (2015) used non-linear pressure diffusion and rock proper­
ties to estimate the seismic triggering front (Eq. 2), quantity and
magnitude, it was assumed that the triggering probability of induced
seismicity was directly related to the pore pressure perturbation:
√̅̅̅̅̅̅̅̅̅̅
Rt = 4πDt (2)

where Rt is the triggering front (the radial distance of the seismic events
from the injection point). Segall and Lu (2015) eliminated the state
variable in Dieterich rate- and state-dependent frictional theory, and
obtained the induced seismicity rate γ from the seismicity state variable
which depends on the change of Coulomb stress:
⎛ ⎞
dγ γ ⎝ τ̇0
= − γ⎠ (3)
dt ta τ̇tec

where γ is the Seismicity Enhancement Factor, it represents the seis­


micity rate at corresponding stressing rate, τ0, and τtec is the tectonic Fig. 6. The diagram of hydraulic diffusivity (D) in the analysis of induced
shear stress rate, ta is the characteristic decay time. Through observation seismicity, the hydraulic diffusivity (D) and the seismic hydraulic diffusivity
and analysis of a wide variety of induced seismicity, Talwani and Acree (Ds) are successively proposed to estimate the magnitude of injection induced
(1985) estimated the possible magnitude of induced seismicity with earthquakes. (modified after Talwani and Acree (1985)).

Fig. 5. The diagram of fault activation associated with CO2 injection, these figures show the pore pressure perturbation process and its effects on fault mechanical
properties and stability, the fault will be activated when the Coulomb stress decreases to a critical value: (a)-(c) are the change of pore pressure multiplied by the
friction coefficient (fΔp). As the injection time increases, the injected CO2 and pore pressure will diffuse along the fault. (d)-(f) are the changes of normal stresses (σn),
the CO2 injection cause the decrease of normal stresses along the fault. (g)-(i) are the changes of shear stresses (τs), pore pressure accumulation in the faults leads to
pore pressure gradients outward from the fault zone, which must be compensated by stress gradients. The combination of the volumetric expansion within and near
the fault zone increases the shear stress on the faults. (j)-(l) are the changes of Coulomb stress (τs + fσ n), the CO2 injection decreases the Coulomb stresses along the
fault, which may reactivate the fault (modified after Chang and Segall, 2016).

8
Y. Cheng et al. Earth-Science Reviews 239 (2023) 104369

diffusivity (D) in analyzing injection induced seismicity. These diffusion contrast to water injection, the viscosity of CO2 is one order of magni­
equations well explain induced seismicity spatial-temporal distribution, tude lower than that of brine; then the CO2 can easily flow into the
and seismic moment. reservoir, and the overpressure usually peaks at the beginning of in­
Table 3 summarizes the recent studies on fault activation related to jection, and then becomes relatively constant (Vilarrasa et al., 2019). In
pore pressure perturbation in CGS. The study of Cheng et al. (2013) addition, fault activation also requires a high permeable fault zone,
showed that injection induced seismicity was mainly caused by pore which is conducive to the diffusion of pore pressure in the fault zone
pressure perturbation; since the precondition for fault activation is that (Yadav et al., 2017). And in other fluid injection projects, such as EGS
the fault has been in a critical state, and the contribution of stress state and wastewater injection, a large number of post injection induced
change caused by pore pressure perturbation to fault activation is usu­ seismic events are observed, and even the maximum magnitude events
ally <3%; that is, pore pressure perturbation could not cause large-scale occur in the post injection stage (Ellsworth, 2013; Rathnaweera et al.,
changes in elastic load and stress state. Generally speaking, the pertur­ 2020). This phenomenon is also worth further consideration for the
bation pressure of water injection gradually increases with time; in induced seismicity estimation of CGS.

Table 3
4.2. Temperature alteration
The studies on fault activation related to pore pressure perturbation.
Reference Coupling condition Reservoir and fault Modeling The temperature of injected CO2 in CGS is lower than that of the
characteristics platform
reservoir. The interaction between the high-temperature reservoir and
Chang and Full poroelastic HM Homogeneous formation, COMSOL cold CO2 plume will cause rock contraction, which will induce thermal
Segall coupling, including isolated/connected
stresses and affect effective stresses. The change of effective stresses will
(2016) pore pressure homogeneous fault
diffusion, Mohr- alter regional stress field, produce thermoelastic strain and thermal
Coulomb stress state fractures (Vilarrasa and Rutqvist, 2017). The non-equilibrium stress
change and rate-state state can lead friction loss and fault slip, this means that the non-
dependent seismicity isothermal effect could affect the fault activation and seismic activity
Rutqvist HM coupling Mohr- Homogeneous poro- TOUGH- (Vilarrasa, 2016; Vilarrasa et al., 2016). The influence region of thermal
et al. Coulomb failure with elastic media, elasto- FLAC stress is controlled by the temperature alteration region induced by CO2
(2011, poro-elasticity plastic fault
cold injection. The perturbation range of the stress is wider than that of
2013)
Guglielmi HM coupling, Discrete fractures 3DEC the hydraulic propagation, then this non-isothermal effect area should
et al. including the embedded in an be larger than that of CO2 plume perturbation. In addition, since thermal
(2008) evaluation of the impervious rock matrix propagation depends on the cooling action of the injected fluid on sur­
shearing effects, CO2 rounding rock, the thermal effect propagation is usually slower than the
plume migration
process and rock
pressure propagation and the hydraulic transmission (Kim and Hosseini,
matrix deformation 2015). Therefore, for several decades CO2 injection, the area of rock
Babarinde HM coupling, non- Homogeneous porous Landmark formation temperature obvious alteration caused by CO2 cooling only
et al. isothermal two-phase media, the identification Nexus occurs within a few hundred meters from the injection well. Although
(2021) (CO2 and brine) flow of fault activation is
compared with the range of CO2 plume, which could reach several
simulation based on Coulomb failure
function square kilometers, the temperature alteration area is very small, the
Schwab et al. Uncoupled, only flow A fault zone with Petrel™ thermal stress can affect the far well stress field through stress transfer
(2017) condition multiple small faults, the and rock contraction deformation (Vilarrasa and Rutqvist, 2017).
fault activation is based Especially when the thermal expansion coefficients of the two forma­
on Andersonian fault
failure criterion
tions are different, and the fault passes through the caprock, the geo-
Urpi et al. HM coupling Describing the fault static TOUGH- mechanical effects caused by cooling may lead to the instability of
(2016) and dynamic friction in FLAC fractures and faults (Vilarrasa and Laloui, 2016).
different homogeneous Vilarrasa et al. (2016) used the linear thermoelastic theory of porous
and heterogeneous
media to analyze the influence of cold CO2 injection on rock strain,
conditions
Cao et al. HM coupling Homogeneous formation, ECLIPSE- establish the thermal-mechanical (TM) stress coupling relationship:
(2021) iteratively method analyzing the fault shear FLAC
σ T = K αT ΔT (5)
slippage and tensile
aperture
Rinaldi et al. HM coupling A double fault system, TOUGH- where K is the bulk modulus, αT is the linear thermal expansion coeffi­
(2015) large tectonic faults FLAC cient of the porous media. Although analytical and semi-analytical
Mazzoldi HM coupling Homogeneous isothermal TOUGH- methods have been developed to analyze the thermal stress caused by
et al. porous media, analyzing FLAC
temperature alteration due to CO2 injection (Bao et al., 2014; LaForce
(2012) undetected small faults
activation et al., 2015), only the initial stage of injection can be well estimated.
Verdon et al. HM coupling Heterogeneous reservoir MORE- When CO2 migrates far away from the injection well, the viscous force
(2011) and homogeneous ELFEN-MPI plays an important role, and the cooling front caused by CO2 plume
caprock and basement cannot be well estimated through the analytical or semi-analytical
Verdon et al. Uncoupled, only flow Reconstruct reservoir and CMG-
(2015) condition faults model from seismic STARS
method. Therefore, for this thermo-hydro-mechanical (THM) coupling
reflection data, discrete problem, the numerical simulation method should be more appropriate
fracture networks, and could obtain more accurate results.
Kroll et al. HM coupling Homogeneous reservoir, NUFT- The influence of thermal stress caused by temperature alteration on
(2020) the fault activation is RSQSim
fault stability also depends on the stress state, i.e. normal fault (vertical
based on rate- and state-
friction, heterogeneous stress is the maximum principal stress), strike slip fault (vertical stress is
stress distribution along the intermediate principal stress) or reverse fault (vertical stress is the
the faults minimum principal stress). Compared to normal and reverse fault, the
Morris et al. HM coupling Homogeneous formation NUFT- activation of strike slip fault is easier to propagate to the caprock
(2011) GEODYN-L
(Vilarrasa and Rutqvist, 2017). Vilarrasa et al. (2015) simulated CO2

9
Y. Cheng et al. Earth-Science Reviews 239 (2023) 104369

injection non-isothermal effects in Salah, Algeria, results showed that principal effective stresses are both horizontal, the changes of the two
the stress field change caused by this temperature alteration triggered horizontal stress caused by overpressure are the same. The size of Mohr’s
shear and tensile failure in the lower part of the caprock. For the normal circle remains unchanged and the circle only moves to the left (HM
faults, the cooling action causes rock contraction, leads to the stress circle). The stress regime becomes less stable than the aforementioned
reduction in all directions, which is proportional to the rock stiffness, normal fault. When considering cooling effect, as the situation of normal
thermal expansion coefficient and temperature change. The cooling re­ fault, the thermal stress causes the same change in stress field. The
gion distributes around the injection well, and is obviously smaller than Mohr’s circle size increases and moves to the yield surface (THM circle),
the extension range of CO2 plume. Due to the local discontinuity of which may lead to fault activation. However, due to the difference in
vertical stress and inelastic strain between the reservoir and the caprock, rock properties and the discontinuity of stresses, the caprock (Fig. 7 d)
the cooling makes the vertical stress of the reservoir less than the weight may be more stable than the reservoir (Fig. 7 c). In the reverse fault
of the caprock (Vilarrasa and Laloui, 2015). In order to satisfy stress situation, the maximum effective stress is horizontal and the minimum
equilibrium and displacement compatibility, the stress redistributes effective stress is vertical. When only considering overpressure, the
around the cooling area, forming an arch effect to support the weight of vertical total stress remains unchanged, horizontal total stress increases.
the caprock. Finally, the total horizontal stress in the lower part of the The size of Mohr circle increases, and the fault is closer to the shear
caprock increases, tightens the caprock, and improves the caprock sta­ failure condition (HM circle). But the impact of thermal effect is less
bility (Vilarrasa et al., 2013). The horizontal stress in the reservoir de­ significant than that in the other two situations. Since the thermal stress
creases, the fault in the reservoir may reach the failure condition (Bao variations are the same in all situations, but the maximum effective
et al., 2014). A similar situation also occurs on the reverse fault in the stress in the reverse fault stress state is much larger, hence the relative
cooling region. Since the maximum principal stress is horizontal, the change of the effective stress caused by TM effect is smaller. Neverthe­
horizontal stress at the lower part of the caprock also increases due to less, the cooling still makes the Mohr circle slightly closer to the yield
cooling action (Vilarrasa et al., 2014), and the reverse fault is generally surface (THM circle), and therefore the stability is also reduced.
under high confining pressure, so the possibility of fault activation Table 4 summarizes the recent studies about THM coupling on the
caused by cooling is small (Vilarrasa et al., 2019; Vilarrasa and Rutqvist, applications of fault activation and induced seismicity in CGS. It is found
2017). Fig. 7 presents the same CO2 injection conditions for normal that the quantification of the injected CO2 cooling effect is very difficult
fault, strike slip fault and reverse fault (Vilarrasa, 2016). The Mohr circle due to the complex thermal effect factors, which include CO2 plume flow
variation trend in these situations could more directly reflect the influ­ characteristics, in-situ stress state, formation heterogeneity and field
ence of non-isothermal effects on fault activation. In the normal fault operating condition. In addition, for most cases, the stress field change
situation, the overpressure reduces the effective stress, the Mohr’s circle caused by cooling action is negligible compared with pressure pertur­
shrinks and moves towards the yield surface (HM circle). The CO2 bation. This thermal stress change just results in shear and tensile fail­
cooling action reduces the thermal stress, and leads to the size of Mohr’s ures of small fractures, induces small magnitude microearthquakes
circle increases and the circle further moves leftward, and the stress state (Vilarrasa and Laloui, 2015). In addition, a high temperature reservoir
is closer to the shear failure condition (THM circle). In the cooling re­ will reduce the CO2 viscosity, promote CO2 flow and mitigate the
gion, for short-term injection, although the stress redistribution can overpressure phenomenon (Wiese et al., 2010). Temperature also affects
offset the change of thermal stress (less than one year); for the long-term the surface tension and wetting angle, which in turn affects the capil­
injection, the change of thermal stress will dominate the change of Mohr larity (Singh et al., 2011) These phenomena imply that the thermal ef­
circle. In the strike slip fault situation, the maximum and minimum fects caused by CO2 should be considered (Class et al., 2009). Therefore,

Fig. 7. The Mohr circle variation trend due to cold CO2 injection: (a), (c) and (e) are the faults in the reservoir; (b), (d) and (f) are the faults in the reservoir. (a) and
(b) are the normal fault, the Mohr’s circle shrinks and moves towards the yield surface (HM circle); the CO2 cooling action reduces the thermal stress, and leads to the
size of Mohr’s circle increases and the circle further moves leftward, and the stress state is closer to the shear failure condition (THM circle). (c) and (d) are the strike
slip faults, the size of Mohr’s circle remains unchanged due to the same change in two horizontal stresses and the circle only moves to the left (HM circle); the CO2
cooling also cause the Mohr’s circle size increases and moves to the yield surface (THM circle). (e) and (f) are the reverse faults, the vertical minimum effective stress
remains unchanged, horizontal maximum effective stress increases, the size of Mohr circle increases (HM circle); the cooling still makes the Mohr circle slightly closer
to the yield surface (THM circle) (modified after Vilarrasa, 2016).

10
Y. Cheng et al. Earth-Science Reviews 239 (2023) 104369

Table 4 evidence that geochemical reaction will cause significant the changes in
The studies on fault activation related to non-isothermal effects. geomechanical properties, and no fault activation induced by
Reference Coupling condition Reservoir and fault Modeling geochemical effects has been observed in existing CGS projects. How­
characteristics platform ever, this issue should not be ignored. The geochemical reaction caused
Kroll et al. The THM coupling Reservoir consist of low- NUFT and by long-time CO2 injection in carbonate rich reservoirs should be
(2020) based on rate- and permeability units, fault RSQSim continuously studied.
state-friction theory, consist of high
including the permeability units 5. Maximum magnitude prediction of CO2 injection induced
variations of
geothermal gradient
seismicity
and CO2 plume
Figueiredo THM coupling, Homogeneous poro- TOUGH- Up to now, there is still no reliable prediction method for injection
et al. include the elastic media, elasto- FLAC- induced seismicity (Cornet, 2015). The main inducement of this kind of
(2015) temperature of plastic fault, fault ECO2N
seismicity is the change of regional stress field caused by fluid injection
formation and CO2 reactivation analysis
Rutqvist THM coupling, non- Homogeneous TOUGH- (Elsworth et al., 2016). Since the fluid injection is controllable, the
et al. isothermal, formation, anisotropic FLAC temporal and spatial distribution of the seismic events is determined by
(2011; multiphase fluid flow elastic fault zone fluid diffusion and the change of stress field, and the large magnitude
2012) events are often related to the fault activation caused by injection
Rinaldi et al. THM coupling Heterogeneous TOUGH-
(2014a, reservoir, analyzing the FLAC
(Foulger et al., 2018). Therefore, this induced seismicity should be
2014b) deformation and slip of theoretically predictable. In recent years, some researchers have studied
heterogeneous fault this problem, including statistical method, physical parameter and
zone simulation method, and hybrid prediction method (McGarr, 2014;
Park et al. THM coupling, the Homogeneous COMSOL-
Rutqvist et al., 2016; Shapiro et al., 2011; Verdon and Stork, 2016). The
(2022) low temperature formation, vertical fault ABAQUS-
supercritical CO2 and plane, analyzing CODE- statistical method is established mainly based on the seismic records
geothermal gradient fracture generation and BRIGHT statistical relationship before and during fluid injection, including the
are considered propagation, simulating time, location and magnitude of seismic events. When using the physics-
seismic wave signal based prediction method, it is necessary to establish a reasonable
propagation
geological model according to the formation physical parameters, and
Dempsey THM coupling, Homogeneous FEHM
et al. volumetric thermal formation, only elastic calibrate the model according to the seismic records and geophysical
(2014) expansion is deformation is data. In this magnitude prediction process, the fluid injection pertur­
considered considered bation, stress temporal and spatial changes, fault activation and earth­
Vilarrasa THM full coupling, Homogeneous CODE-
quake triggering should be characterized. The hybrid prediction method
et al. non-isothermal CO2 formation, analyzing BRIGHT
(2016) injection the normal, reverse and includes not only seismic statistical analysis, but also reservoir geolog­
strike slip faults ical characteristics and injection perturbation description.
activation Among all statistical theories, the statistical model based on injection
volume proposed by McGarr (2014) is generally accepted. This theory
assumes that the fully saturated the rock formation contains potential
more comprehensive studies are needed to reveal the influence of tem­
activated faults near the injection well in the influence range of stress
perature alteration on CO2 injected induced seismicity.
field variation. The variation upper limits of the regional stress field and
the effective stress on the fault are determined by the injection volume,
4.3. Geochemical effect it means that the seismic events are also limited by the injection volume.
Therefore, the linear relationship between the maximum magnitude and
The long-term injection of CO2 into the reservoir will lead to the injection volume can be given based on Gutenberg-Richter earth­
chemico-mineralogical changes in rock matrix and fault. On one hand, quake distribution law.
the stress-aided corrosion and chemico-mineralogical effect lead to the
Mmax = GΔV (6)
alterations of friction and cohesion. This friction loss will perturb the in-
situ stress conditions and even result in fault activation (Jia et al., 2020). where G is the modulus of rigidity, ΔV is the injection volume.
On the other hand, the dissolution of CO2 into brine will form an acidic McGarr compared Eq. 6 with the maximum magnitudes of major
solution, which has the potential to dissolve minerals. This in turn may injection projects in recent decades, including scientific research, hy­
result in the subsequent precipitation of other minerals (Zhang et al., draulic fracturing, geothermal and wastewater disposal. His research
2009). This mineral alteration will lead to friction and cohesion insta­ results showed that this upper limit could arrest the magnitudes in most
bility, reduce surface free energy and formation strength, and create injection projects. However, the maximum magnitude in Pohang injec­
new fractures (Rathnaweera et al., 2020). In addition, in carbonate tion project violated this the upper limit of the magnitude. The reason is
reservoir and carbonate rich mudstone, the geochemical reaction is the that this theory lacks the consideration of hydrogeological conditions.
fastest (Vilarrasa et al., 2019). Carbonate mineral reacts with acid so­ Shapiro et al. (2010) derived a statistical parameter (seismogenic
lution, resulting in the increase of rock porosity and permeability, which index) by analyze the activity rate of induced seismicity and fluid in­
in turn leads to the decrease of rock stiffness and strength, and affects jection rate, the seismogenic index represents the seismic tectonic state
fault activation (Alam et al., 2014). The measurement results in the near the injection well. The seismogenic index provides another statis­
laboratory show that the rock strength can be reduced by about 20%– tical method to predict the induced seismicity. This index is determined
30% (Kim et al., 2018b; Vialle and Vanorio, 2011). by reservoir tectonic characteristics and independent of the injection
The reaction of rock with CO2 and brine depends on the formation parameters, every site should has a specific seismogenic index, which
type, and carbonate and feldspar minerals are more easily dissolved in does not change with time. (Dinske and Shapiro, 2013). It can be used to
acidic environment. However, laboratory studies show that the effects of quantitatively evaluate the probabilities of seismic magnitudes in fluid
geochemical processes on geomechanical properties are usually minor injection projects. The magnitude estimation is also based on Gutenberg
(Rohmer et al., 2016). The observation of fault exposed to acidic con­ and Richter earthquake distribution law.
ditions for a long time in the natural environment also authenticate
phenomenon (Bakker et al., 2016). Up to now, for CGS, there is no

11
Y. Cheng et al. Earth-Science Reviews 239 (2023) 104369

logNM (t) = logQ(t) − bM + Σ (7) where γ represent the reservoir physical characteristics, it is determined
by dynamic friction coefficient, the background stress drop, the forma­
where M is the magnitude, NM represents the seismic quantity, b is the
tion bulk modulus and reservoir size. Compared with McGarr’s statis­
Gutenberg-Richter exponent, Q is the cumulative injected volume and Σ
tical method, this method adds the analysis of the physical process of
is the seismogenic index.
earthquake nucleation, and hence the accuracy of the maximum
The large seismogenic index indicates that the probability of large
magnitude prediction can be improved.
magnitude events is high. Shapiro et al. (2010) used the critical pressure
Fig. 8 shows the results of recorded maximum magnitudes in
(C) to define the seismogenic index, the critical pressures represent the
different injection projects and maximum predicted magnitude upper
inherent defects of the reservoir, that is, potential seismogenic sources.
limits of McGarr and Galis et al. Almost all the maximum magnitudes are
The critical pressures are used to simulate the induced seismicity, and
lower than these two magnitude upper limits, and only a few seismic
the pore pressure perturbation is the only triggering factor (Rothert and
maximum events exceed the predicted upper limits. In particular, the
Shapiro, 2007; Shapiro et al., 2010). Even though the reservoir property
Pohang injection induced maximum magnitude seismic event shows
(the formation inherent defect) is considered in this model, the
that the relationship between the maximum magnitude and the injection
maximum magnitude prediction in this method is only related to the
volume is unreliable (Grigoli et al., 2018). It means that the injection
stimulated volume, and the probability of large earthquakes is propor­
behavior can trigger significant earthquakes. Under specific circum­
tional to the volume of pressurized zone, the tectonic factors such as
stances, fault activation may be out of control, and the induced
tectonic deformation and fault activation are ignored. Therefore, this
maximum magnitude can far exceed the result estimated by injection
model may lead to serious magnitude underestimation.
volume (Kim et al., 2018c). That is to say, the injection maximum
Van der Elst et al. (2016) established another statistical maximum
magnitude is not constrained by the net injection volume. Although the
magnitude prediction method based on the reservoir tectonic charac­
current seismic monitoring results of all CO2 injection projects show that
teristics. They consider the injection induced earthquakes and the tec­
the maximum magnitude conforms to these two magnitude upper limit
tonic earthquakes have the same magnitude upper limit for a specific
models, the supercritical CO2 and water have similar effects on fault
reservoirs. Therefore, according to the early observed seismicity, the
activation. Therefore, the CO2 injection may also trigger events that
maximum magnitude of injection induced earthquakes can be predicted:
exceed these two magnitude upper limits.
1 The physical based method is an alternative to statistical method, as
Mmax = Mc + logQ(t) + log10 NM (8)
b another typical physical based method, the numerical simulation
method can be used to simulate reservoir physical evolution process
where Mc is reference magnitude, NM is the number of earthquakes during injection, such as pore pressure perturbation, regional stress field
observed above magnitude Mc, b is the Gutenberg-Richter exponent change, fault slip, etc. (Fig. 9). For CO2 injection project, it is generally
(Van der Elst et al., 2016). In this method, the change of tectonic stress assumed that the induced seismicity occurs on the pre-existing fractures
and the release of strain energy are regarded as the main inducement of or faults (Cappa and Rutqvist, 2011; Rutqvist et al., 2016; Verdon and
seismic nucleation, while the influences of injection pressure perturb­ Stork, 2016; Verdon et al., 2015). Therefore, before simulation, it is
ance and fluid volume on magnitude prediction are ignored. Hence, if necessary to know the location, orientation and size of the fracture/fault
the induced seismicity is caused by fault activation, the maximum
magnitude is determined by the fault tectonic condition. Recent studies
and observations on the maximum magnitudes of injection induced
seismicity also authenticate this view (Gan and Frohlich, 2013b; Grigoli
et al., 2018; Rutqvist et al., 2016). This model also may fail in sesimic
prediction situations. Since this statistical method is based on previous
induced seismicity observation, when small earthquakes accumulate to
a certain number, the prediction results show that large earthquakes
should nucleate; while for stable reservoirs, there may be no large
magnitude event after a long time injection. The results of some CO2 or
other injection projects show that no large earthquakes are induced after
a long time or large volume injection (Keranen et al., 2014; Van der Elst
et al., 2016; White and Foxall, 2016). On the other hand, when the in­
jection pressure or rate increases suddenly, or the stress perturbation
caused by injection activity reaches the tectonic fault after a period of
time, it is possible to trigger large magnitude events suddenly; this
method does not take into account the sudden change of injection
behavior and the heterogeneity of reservoir, and assumes that the seis­
micity triggering is a gradual evolution process; These situations do not
conform to the reality. Therefore, the prediction of nucleation, rupture
and magnitude with this method is still a challenge.
Galis et al. (2017) proposed a hybrid prediction method and try to
improve the phenomenon of magnitude underestimation to some extent. Fig. 8. The recorded maximum injection induced seismic magnitudes and
In this method, the rock breakdown physical process associated with maximum predicted magnitude upper limits of McGarr and Galis et al. The
injection parameters is firstly established; then the rupture size caused yellow solid line is the McGarr’s maximum magnitude prediction model, the
by local pore pressure perturbation is theoretically estimated, and the black dashed line is the Galis’s maximum magnitude prediction model. The
triangles are the maximum magnitudes in CGS, EGS, wastewater disposal, sci­
propagation of stress field along the fault is predicted; at last, the
entific research projects. The red triangle is the maximum magnitude in Pohang
theoretical estimation between maximum magnitude and injection
EGS project., the other colour triangles are the maximum magnitudes in CGS
volume is established. The theoretical estimated largest magnitude can projects. Compared with the McGarr’s model, the Galis’s model has an
be expressed as: improvement on magnitude underestimation, which arrests all maximum
magnitudes except for the Pohang maximum magnitude event. (For interpre­
Mmax = γ • ΔV 3/2 (9)
tation of the references to colour in this figure legend, the reader is referred to
the web version of this article.)

12
Y. Cheng et al. Earth-Science Reviews 239 (2023) 104369

characterization and seismic occurrence rate, the PSHA infers a hazard


curve, which represents the predicted probabilities of exceeding ground
motion values. The classical PSHA is established to predict tectonic
earthquake hazards. The impact of injection activities on tectonic faults
is not fully considered, hence the risk of large earthquakes could not be
reasonably predicted under specific circumstances (White and Foxall,
2016). Therefore, it is difficult to put forward practical guidelines when
evaluating the probability of seismic activities induced by CO2 injection
(Pawar et al., 2015). The traffic light system is designed to forecast the
probability of induced seismicity based on the early monitoring infor­
mation. Ideally, seismicity observation can indicate that previously
unknown faults which are encountered by the injected fluid, and the
potential information can be used to revise the later seismicity fore­
casting and revise the injection strategy to avoid potential larger
magnitude events. However, there may be almost no early warning
before the occurrence of a significant earthquake. In this case, the traffic
light system will fail. The M3.4 earthquake in the Basel geothermal field,
Switzerland and the M5.5 event in the Pohang geothermal operation,
South Korea were triggered in this situation, and the traffic light system
Fig. 9. The simulation results of fault activation induced by CO2 injection: (a)
could not provide useful warning information before the occurrence of
Plastic shear strain along the fault and slip profile at the time of seismic trig­
large magnitude events (Grigoli et al., 2018; Häring et al., 2008; Kim
gering, the white triangles are the control points along the fault; (b) CO2
saturation of the control point 5; (c) fluid pressure of the control point 5. The
et al., 2018a). The CGS projects will sequester a large amount of CO2
simulation results show that the though CO2 is mainly situated near the in­ (megatonnes to gigatonnes), create large overpressured regions, and
jection point, the fluid pressure already migrates to the fault and induces an could activate the large tectonic faults, trigger significant events
abrupt stress perturbation and subsequent fault slip (modified after Rutqvist exceeding the prediction of traffic light system (White and Foxall, 2016;
et al., 2016). Zoback and Gorelick, 2012). The aforementioned statistical models and
numerical simulation also encounter various challenges in the actual
in the reservoir, the initial stress conditions at the beginning of injection seismic activity forecasting (Rathnaweera et al., 2020; Rutqvist et al.,
and the stress field change process during injection. Then, seismic ac­ 2016; White and Foxall, 2016).
tivity can be determined by simulating when and where the stress on the In CGS operations, one of the difficulties is to evaluate and explain
fracture/fault exceeds the Mohr Coulomb failure strength. Verdon and the interaction between CO2 plume and faults. In order to analyze the
Stork (2016) used the this method to predict the maximum magnitude of seismic activity induced by CO2 injection, it is necessary to consider the
the In Salah CO2 injection project, the predicted result was in good stress field changes caused by many reasons, such as pore pressure
agreement with the recorded largest event, which had a magnitude of perturbation, non-isothermal effect, chemical mineral dissolution and
1.7. precipitation, as well as the coupling of these processes. To forecast the
Both the physical model and the statistical model have made induced seismicity, it is also necessary to better understand the complex
reasonable work on the prediction of injection induced maximum fault system and tectonic stress field. However, it is still impossible to
magnitude. The statistical model could predict the maximum magnitude accurately determine how the complex fault system affects the injection
without geological information. The main advantage is that it is easy to induced seismicity and the maximum magnitudes. In addition, the
use. The seismicity rate and b value can be calculated by seismic records, hidden tectonic faults, that may be located outside the reservoir, have
then the maximum magnitude can be predicted. However, the stability great risk and uncontrollability when large-scale CO2 are injected un­
of b value requires the stability of injection behavior and the homoge­ derground (Zoback and Gorelick, 2012).
neity of geological formation. For the sudden change of geological Technically, reasonable injection strategies can mitigate the indued
structure encountered in the process of pore pressure diffusion, the seismic hazard in CGS. Different from tectonic stress field, the CO2 in­
statistical methods are difficult to accurately predict the maximum jection can be controlled, including injection rate, injection volume,
magnitude. In contrast, the physical model needs to provide a lot of injection pressure, etc. So far, several injection induced seismicity risk
geological information. The prediction results of the physical method mitigation tests have been carried out, the intermittent injection method
will completely depend on the input geological parameters and engi­ and shear reservoir stimulation scheme are the most commonly used
neering parameters, including well logs, three-dimensional reflection approaches to mitigate induced seismicity (Hofmann et al., 2019;
seismic imaging, rock physical property testing, injection rate, injection Kwiatek et al., 2019; Zang et al., 2013). These methods aim to alter the
pressure, injection volume, etc., which requires a lot of manpower and stress release process so as to transform potential large earthquakes into
financial resources to ensure the reliability of the prediction results. multiple small magnitude ones (Cheng et al., 2022; Kim et al., 2018a).
However, these seismic mitigation methods fails in the Pohang EGS
6. Forecasting challenges of CO2 injection induced seismicity project (Grigoli et al., 2018; Kim et al., 2018c), this means that the
effectiveness of these approaches need to be further verified, improved
The induced seismic monitoring results of on-site injection projects and studied.
in the past decades indicate that although various methods have been The complexity of triggering mechanism and influence factors of
adopted to reduce the possibility of significant earthquakes, the large injection induced seismicity makes the implementation of actual fore­
magnitude events cannot be completely avoided (Cornet, 2015; Lee cast challenging. In the view of the aforementioned critical issues, we
et al., 2019; McGarr et al., 2015). Therefore, more accurate and reliable summarized the forecast challenges of induced seismicity related to
methods are needed to forecast the induced seismicity risks. So far, some CGS.
strategies for controlling and mitigating the injection induced seismicity
have been developed, including probabilistic seismic hazard assessment 6.1. Accurate characterization of stress state
(PSHA), traffic light systems, statistical analyzing forecasting methods
and numerical simulation models. On the basis of seismic source In order to better forecast the seismic activity induced by CO2 in­
jection in CGS, it is important to identify the initial stress state and

13
Y. Cheng et al. Earth-Science Reviews 239 (2023) 104369

accurately characterize the change of stress process during injection. in 3D seismic survey and interpretation. Only the faults over 5 km are
The stress field will directly affect the distribution of CO2 plume, thus easy to identify. But in most CGS projects, the M2 earthquake is already
affecting the temporal and spatial of induced seismicity and the stability unbearable. This means that in most CO2 injection projects, many po­
of faults. Especially, the correct quantification of principal stresses are tential seismic risks are unpredictable. In the field fault survey, the
particularly important, which will directly affect the analysis of rock detection resolution is crucial. Only by accurately interpreting the fault,
fracture direction, fluid seepage, pore pressure diffusion and local stress can a reasonable injection scheme be designed. Therefore, although the
change around the faults. However, accurate characterization of the potential risks of large faults is higher, the overall risks may be easier to
stress state is still a very challenging task in CGS. manage since they are easy to identify. However, small faults that are
difficult to identify may cause unexpected seismic risks. It is the
complexity of fault system that limits the forecasting of injection
6.2. Correct identification of fault systems induced seismicity.

For CGS, reservoirs with high permeability are generally selected, so


the induced earthquakes, especially large magnitude events, are often 6.3. Fluid injection scheme
associated with faults (Rutqvist, 2012; Rutqvist et al., 2016; White and
Foxall, 2016; Zoback and Gorelick, 2012). Therefore, the faults (or fault The CO2 injection will cause the continuous expansion of pressure
systems) correct identification and characterization becomes the key for perturbation. Therefore, the induced seismicity has a strong temporal
the forecasting of induced seismicity. Geophysical detection technolo­ and spatial dependence on the injection scheme, leading to the temporal
gies have inherent limitations, which makes the faults identification and spatial evolution of induced seismicity with the pressure perturba­
(especially small faults) difficult. In general, pre-existing faults near the tion. At present, although the established CGS projects design CO2 in­
injection projects can be divided into known faults and unknown faults. jection schemes, more extensive researches are needed to fully
The problems relate to the known faults can be solved in the early understand the impact of injection characteristics on the induced seis­
design stage of a project. The project can be designed far away from the micity. Due to the lack of understanding of how different injection
fault to avoid the fault activation risk. For low-risk faults, the possibility scenarios (including the changes of injection rates, pressures and vol­
of reactivation should be evaluated first, then careful seismic monitoring umes) perturb critical stressed faults, it limits the safe and effective
needs to be carried out during CO2 injection, and corresponding fault implementation of CGS.
activation mitigation plan should be developed. In fact, some CGS pro­ Up to now, there are two main methods to predict the impact of
jects are established close to large tectonic faults, such as the Snøhvit in injection on induced seismicity: empirical method and simulation
Norway (Chiaramonte et al., 2015) and the Otway in Australia (Vidal- method. The empirical method predicts the seismic hazards based on the
Gilbert et al., 2010). These faults have been considered to be low-risk in previous seismic records. As the injection proceeds, more and more
the early exploration and project design stage, and the on-site seismic seismic events could be used to continuously recalibrate the parameters
monitoring also reveals a low level of induced seismicity during CO2 of the forecasting model. These empirical methods usually simplify the
injection. basic physical processes, the forecasting parameters depend on the
In fault identification, the accurate identification of fault size is the observed seismic data. Therefore, the results of this method will depend
key to fault detection. The size of the fault determines the rupture range on (1) the fidelity of the simplified model to the real physical processes,
during fault activation, which is directly related to the potential induced and (2) sufficient recorded seismic events. These two problems have not
seismic magnitude. Previous studies (Hanks, 1977; Scholz, 2019) have been solved yet, and the reliability of the forecasting system cannot be
shown that a fault of 100 m could trigger an M2 earthquake (Fig. 10), guaranteed at present. The recent developments of empirical methods
which could be felt by the residents near the project. An 1 km fault could have been applied to the water injection in EGS project (Bachmann
trigger M3 to M4 earthquakes (Fig. 10), which could cause minor et al., 2011; Hajati et al., 2015; Hummel and Shapiro, 2016; Mena et al.,
damage to infrastructure and even cause light casualties. The faults 2013; Rothert and Shapiro, 2007). Even though some promise is shown,
below 100 m are usually indetectable in geological survey, and the fault there are still engineering failure cases for the forecasting of induced
with the length of 1 km is also difficult to identify, which may be missed earthquakes. In addition, unlike EGS reservoir stimulation, CGS projects

Fig. 10. The relationship between fault length and


induced seismic magnitude, black dashed lines
represent potential magnitude range during fault
activation, vertical shaded regions indicate the visi­
bility of a given size fault using geophysical detection
means, the red solid line represent the magnitude = 2,
which is an common lower limit of a felt earthquake.
The undetectable faults (the invisible and challenging
zones) could cause unbearable seismic events (M ≥ 2)
(modified after Scholz, 2019; White and Foxall,
2016). (For interpretation of the references to colour
in this figure legend, the reader is referred to the web
version of this article.)

14
Y. Cheng et al. Earth-Science Reviews 239 (2023) 104369

operate in high permeability reservoirs for a long time. What extent the and should be controllable. However, the large earthquakes observed in
EGS seismic forecasting model could be transformed into CGS opera­ the some injection projects indicate that it is still challenging to select a
tions is worthy of further study. reasonable seismic catalog and assessment workflow to forecast seismic
The numerical simulation has been applied to the induced seismicity hazard and guide the subsequent injection activities. Due to the limi­
forecast in CGS, EGS and wastewater injection operations (Baisch et al., tation of geophysical detection technologies, there are always some
2010; Cappa and Rutqvist, 2011; Dieterich et al., 2015; Figueiredo et al., unknown faults in the reservoir, such as the Decatur CGS project in the
2015; Rutqvist et al., 2011; Vilarrasa et al., 2019; Vilarrasa and Rutqvist, Illinois basin. Microseismic interpretation results show that there are a
2017). These models can simulate fracturing process or friction failure large number of unknown small faults near this project. It will be diffi­
under different complex injection situations. Theoretically, the numer­ cult to assess the risks of earthquakes induced by these unknown faults.
ical model is closer to the real injection physical evolution process than
the empirical model, and can simulate the seismogenic process in any
complex situation. However, in practice, it is still a big challenge to 6.5. Recent new technology
obtain accurate characterization data to calibrate the numerical model.
It is difficult to precisely identify the reservoir geological structure
due to the complex and changeable underground conditions. Therefore,
6.4. Seismic risk assessment it is a challenging work to monitor and characterize the CO2 plume
underground migration. Fig. 11 summarizes the current forecasting
Natural seismic hazard assessment has been widely studied, a lot of methods which could be used in the CO2 injection induced seismicity,
practical work can be used as references to solve the current injection and we still face great challenges in these studies. Recently, some new
induced seismicity problems. The real difficulty of this approach is to achievements may help us improve the abilities to forecast the CO2 in­
ensure that the selected assessment workflow is reasonable. The classical jection induced seismic activity, such as the ground penetrating radars
probabilistic seismic hazard assessment (PSHA) method needs to char­ and optical fiber sensors in seismic monitoring. These detection methods
acterize and identify the seismogenic geology structure. In addition, it is may be able to more accurately monitor the triggering process of
required to input the seismic event catalog, estimate the seismic fre­ induced seismicity and help us reasonably predict future seismic risks.
quency, and obtain the correlation between the seismogenic source and However, these methods are still in in their infancy, and the field
time through the seismic occurrence rate. However, the classical PSHA application still faces many obstacles. In addition, Internet of Things
method is based on natural seismicity. Natural earthquakes generally (IoT) and deep learning methods have recently attracted the more at­
have a long historical catalog, which can be used to identify the seis­ tentions due to their abilities and efficiency to deal with complex issues
mogenic sources and estimate the long-term recurrence rates. Moreover, (e.g., geological hazard forecasting, environmental risk assessment).
the natural earthquake frequency is often stable in time. However, the Fig. 12 shows the conceptual diagram of injection induced seismicity
injection induced seismicity depends on the temporal and spatial vari­ monitoring and controlling system based on the IoT and the deep
ations related to injection behaviors, this means that the natural seismic learning technology. This system can monitor the changes of earth­
records before injection are often unavailable. Therefore, new methods quakes, displacements, temperatures and pressures in the reservoir
are needed to evaluate the potential seismicity and its relationship with during injection process in real time, and can measure, capture and
injection activities. The occurrence rates of natural earthquakes are transmit these information in real time. The deep learning methods may
basically determined by the tectonism which is beyond the control of also further improve the accuracy of induced seismicity forecasting.
human beings. Therefore, the risk mitigation is usually focused on site However, the practical application of this system still faces great chal­
selection and injection scheme design. Correspondingly, occurrence lenges, including the deployment of monitoring instruments in deep and
rates of injection induced seismicity are related to injection behaviors, complex underground environments, high costs, accurate monitoring

Fig. 11. The forecasting methods of injection-induced seismicity in CGS (modified after Gaucher et al., 2015).

15
Y. Cheng et al. Earth-Science Reviews 239 (2023) 104369

Fig. 12. The conceptual diagram for injection induced seismicity monitoring system in CGS based on Internet of Things and deep learning method (modified after
Rathnaweera et al., 2020).

and acquisition of data in extreme environments, and general deep great challenge to predict the activation risk of unknown faults.
learning algorithms that can be applied to injection induced seismicity. 3. Pre-existing faults in critical stress state are more sensitive to fluid
These factors limit the applications of these advanced technologies. perturbation. Pore pressure diffusion caused by CO2 injection, thermal
stress caused by temperature alteration, and stress corrosion caused by
7. Conclusion long-term chemical action will have serious impacts on the faults under
the critical stress state. Although some studies have been carried out on
The purpose of this paper is to elaborate the core issues about CO2 fault activation under THMC coupling, it is still a challenge to apply
injection induced seismicity, focusing on the influences and applications these THMC coupling processes on field fault activation assessment and
of THMC coupling on the CO2 injection induced seismicity. Although analysis, correlate temporal and spatial distribution of induced seis­
these efforts are encouraging and have made some progress, there are micity with fault activation, predict the maximum magnitudes, etc. In
still some obstacles that need to be addressed urgently. Therefore, we order to respond to these challenges, the reservoir model under THMC
summarized the issues faced in the studies of CO2 injection induced coupling can be mixed with the seismic triggering simulation to improve
seismicity as follows: the prediction of injection induced seismicity.
1. The induced seismicity recorded in the CGS field operations is 4. Both the physical model and the statistical model have made
summarized, and the basic mechanisms of injection induced seismicity reasonable work on the prediction of the injection maximum magnitude.
are analyzed, including critical pressure theory, Biot’s incremental However, under limited data, the physical model is difficult to accu­
strain theory, rate- and state-dependent frictional theory and fracture rately characterize the reservoir, and the statistical model lacks
potential theory. In these mechanisms, pore pressure perturbation is geological information, which result in challenges for both methods in
considered to be the key of seismicity triggering. the prediction of maximum magnitude. At present, the hybrid prediction
2. Faults are ubiquitous underground. Whereas, the ability to iden­ method based on statistical and physical models can improved magni­
tify and characterize them through geophysical technologies is still tude prediction to some extent, but the problem magnitude underesti­
limited. Especially for moderate-sized faults, they are large enough to mation is still not solved.
generate worrying seismic activity, but small enough to be difficult to 5. Although the CO2 injection induced seismicity is facing challenges
detect by traditional detection technical methods. Therefore, it is still a in fault identification, injection scheme and seismic risk analysis, the use

16
Y. Cheng et al. Earth-Science Reviews 239 (2023) 104369

of new geophysical techniques, IoT and deep learning methods is Chang, K.W., Segall, P., 2016. Injection-induced seismicity on basement faults including
poroelastic stressing. J. Geophys. Res. Solid Earth 121 (4), 2708–2726.
promising to improve the deficiencies of current technologies and re­
Cheng, H., Shi, Y., Zhang, H., 2013. Application of Poroelastic FEM model to the ML5. 7
searches. In addition, the maximum magnitude prediction method based earthquake triggered by the Aswan Reservoir, AGU Fall Meeting Abstracts. T11C-
on deep learning may overcome the shortcomings of traditional 2468.
magnitude prediction methods and obtain satisfactory results in prac­ Cheng, Y., Xu, T., Zhang, Y., Shangguan, S., Feng, B., 2022. An laboratorial investigation
of induced seismicity characteristics in EGS hydraulic fracturing. Geothermics 105,
tical application. 102482.
6. To mitigate potential induced seismicity, the site of CGS should be Chiaramonte, L., Johnson, S., White, J., 2011. Preliminary geomechanical analysis of
selected far away from the large faults, the reservoir with developed CO2 injection at Snøhvit, Norway, 45th US Rock Mechanics/Geomechanics
Symposium. OnePetro.
fracture network is a better choice, then the injection pressure can be Chiaramonte, L., White, J.A., Trainor-Guitton, W., 2015. Probabilistic geomechanical
easier to dissipate. The IoT for seismic monitoring and traffic light sys­ analysis of compartmentalization at the Snøhvit CO2 sequestration project.
tem for seismic prediction should be established to analyze seismic risk J. Geophys. Res. Solid Earth 120 (2), 1195–1209.
Chiaramonte, L., Zoback, M.D., Friedmann, J., Stamp, V., 2008. Seal integrity and
in real time, once large magnitude events may occur, the CO2 injection feasibility of CO2 sequestration in the Teapot Dome EOR pilot: geomechanical site
should be stopped immediately. characterization. Environ. Geol. 54 (8), 1667–1675.
The CGS can significantly help us reduce greenhouse gas emissions. Class, H., Ebigbo, A., Helmig, R., Dahle, H.K., Nordbotten, J.M., Celia, M.A.,
Audigane, P., Darcis, M., Ennis-King, J., Fan, Y., 2009. A benchmark study on
However, the seismicity induced by CO2 injection is a serious problem. problems related to CO2 storage in geologic formations. Comput. Geosci. 13 (4),
At present, the field experience of CGS induced seismicity is still very 409–434.
limited. With the increase of field experience, relevant technologies will Connolly, P., Cosgrove, J., 1999. Prediction of fracture-induced permeability and fluid
flow in the crust using experimental stress data. AAPG Bull. 83 (5), 757–777.
continue to be improved. So far, the implementation of several CO2
Cornet, F.H., 2015. Earthquakes induced by fluid injections. Science 348 (6240),
injection projects also shows that reasonable site selection and careful 1204–1205.
project design can reduce the seismic risk to an acceptable low level. In Couëslan, M.L., Butsch, R., Will, R., Locke II, R.A., 2014. Integrated reservoir monitoring
the future, with the continuous progress of technology, we believe that at the Illinois Basin–Decatur Project. Energy Procedia 63, 2836–2847.
Council, N.R., 2013. Induced Seismicity Potential in Energy Technologies. National
we can safely implement CGS projects. Academies Press.
Daley, T., 2009. Borehole seismic monitoring at Otway using the Naylor-1 instrument
string. California Digital Library, University of California.
Declaration of Competing Interest Damen, K., Faaij, A., Turkenburg, W., 2006. Health, safety and environmental risks of
underground CO2 storage–overview of mechanisms and current knowledge. Clim.
Chang. 74 (1), 289–318.
The authors declare that they have no known competing financial Dance, T., 2013. Assessment and geological characterisation of the CO2CRC Otway
interests or personal relationships that could have appeared to influence Project CO2 storage demonstration site: from prefeasibility to injection. Marine
Petrol. Geol. 46, 251–269.
the work reported in this paper.
Davis, S.D., Pennington, W.D., 1989. Induced seismic deformation in the Cogdell oil field
of West Texas. Bull. Seismol. Soc. Am. 79 (5), 1477–1495.
Data availability Dempsey, D., Kelkar, S., Pawar, R., Keating, E., Coblentz, D., 2014. Modeling caprock
bending stresses and their potential for induced seismicity during CO2 injection. Int.
J. Greenhouse Gas Control 22, 223–236.
Data will be made available on request. Dieterich, J., 1994. A constitutive law for rate of earthquake production and its
application to earthquake clustering. J. Geophys. Res.Solid Earth 99 (B2),
2601–2618.
Acknowledgments Dieterich, J.H., Richards-Dinger, K.B., Kroll, K.A.J.S.R.L., 2015. Modeling injection-
induced seismicity with the physics-based earthquake simulator. RSQSim 86 (4),
This study has been jointly funded by the National Natural Science 1102–1109.
Dinske, C., Shapiro, S.A., 2013. Seismotectonic state of reservoirs inferred from
Foundation of China (No. 42141013), China Postdoctoral Science
magnitude distributions of fluid-induced seismicity. J. Seismol. 17 (1), 13–25.
Foundation (Grant No. 2022 M711291) and the Engineering Research Eggleston, H., Buendia, L., Miwa, K., Ngara, T., Tanabe, K., 2006. 2006 IPCC guidelines
Center of Geothermal Resources Development Technology and Equip­ for national greenhouse gas inventories.
ment, Ministry of Education, Jilin University. Ellsworth, W.L., 2013. Injection-induced earthquakes. Science 341 (6142), 1225942.
Elsworth, D., Spiers, C.J., Niemeijer, A.R., 2016. Understanding induced seismicity.
Science 354 (6318), 1380–1381.
References Evans, K.F., Zappone, A., Kraft, T., Deichmann, N., Moia, F., 2012. A survey of the
induced seismic responses to fluid injection in geothermal and CO2 reservoirs in
Europe. Geothermics 41, 30–54.
Alam, M.M., Hjuler, M.L., Christensen, H.F., Fabricius, I.L., 2014. Petrophysical and rock-
Ferronato, M., Gambolati, G., Janna, C., Teatini, P., 2010. Geomechanical issues of
mechanics effects of CO2 injection for enhanced oil recovery: experimental study on
anthropogenic CO2 sequestration in exploited gas fields. Energy Convers. Manag. 51
chalk from South Arne field, North Sea. J.Petrol.Sci.Eng. 122, 468–487.
(10), 1918–1928.
Aminu, M.D., Nabavi, S.A., Rochelle, C.A., Manovic, V., 2017. A review of developments
Figueiredo, B., Tsang, C.-F., Rutqvist, J., Bensabat, J., Niemi, A., 2015. Coupled hydro-
in carbon dioxide storage. Appl. Energy 208, 1389–1419.
mechanical processes and fault reactivation induced by CO2 injection in a three-
Babarinde, O., Okwen, R., Frailey, S., Yang, F., Whittaker, S., Sweet, D., 2021.
layer storage formation. Int. J. Greenhouse Gas Control 39, 432–448.
A workflow to assess the efficacy of brine extraction for managing injection-induced
Foulger, G.R., Wilson, M.P., Gluyas, J.G., Julian, B.R., Davies, R.J., 2018. Global review
seismicity potential using data from a CO2 injection site near Decatur, Illinois. Int.J.
of human-induced earthquakes. Earth Sci. Rev. 178, 438–514.
Greenhouse Gas Control 109, 103393.
Freifeld, B., Zakim, S., Pan, L., Cutright, B., Sheu, M., Doughty, C., Held, T., 2013.
Bachmann, C.E., Wiemer, S., Woessner, J., Hainzl, S., 2011. Statistical analysis of the
Geothermal energy production coupled with CCS: a field demonstration at the
induced Basel 2006 earthquake sequence: introducing a probability-based
SECARB Cranfield Site, Cranfield, Mississippi, USA. Energy Procedia 37, 6595–6603.
monitoring approach for Enhanced Geothermal Systems. Geophys. J. Int. 186 (2),
Galis, M., Ampuero, J.P., Mai, P.M., Cappa, F., 2017. Induced seismicity provides insight
793–807.
into why earthquake ruptures stop. Sci. Adv. 3 (12), eaap7528.
Baisch, S., Vörös, R., Rothert, E., Stang, H., Jung, R., Schellschmidt, R., 2010.
Gan, W., Frohlich, C., 2013. Did CO2 injection induce 2006-2011 earthquakes in the
A numerical model for fluid injection induced seismicity at Soultz-sous-Forêts. Int. J.
Cogdell oil field, Texas?, AGU Fall Meeting Abstracts, pp. S32C-03.
Rock Mech. Min. Sci. 47 (3), 405–413.
Gan, W., Frohlich, C., 2013b. Gas injection may have triggered earthquakes in the
Bakker, E., Hangx, S.J., Niemeijer, A.R., Spiers, C.J., 2016. Frictional behaviour and
Cogdell oil field, Texas. Proc. Natl. Acad. Sci. 110 (47), 18786–18791.
transport properties of simulated fault gouges derived from a natural CO2 reservoir.
Gaucher, E., Schoenball, M., Heidbach, O., Zang, A., Fokker, P.A., van Wees, J.-D.,
Int. J. Greenhouse Gas Control 54, 70–83.
Kohl, T., 2015. Induced seismicity in geothermal reservoirs: a review of forecasting
Bao, J., Xu, Z., Fang, Y., 2014. A coupled thermal-hydro-mechanical simulation for
approaches. Renew. Sustain Energy Rev. 52, 1473–1490.
carbon dioxide sequestration. Environ. Geotechn. 3 (5), 312–324.
Geneva, 2013. Intergovernmental Panel on Climate Change, 2014, Switzerland.
Bauer, R.A., Carney, M., Finley, R.J., 2016. Overview of microseismic response to CO2
Grigoli, F., Cesca, S., Rinaldi, A.P., Manconi, A., Lopez-Comino, J.A., Clinton, J.F.,
injection into the Mt. Simon saline reservoir at the Illinois Basin-Decatur Project. Int.
Westaway, R., Cauzzi, C., Dahm, T., Wiemer, S., 2018. The November 2017 M-w 5.5
J. Greenhouse Gas Control 54, 378–388.
Pohang earthquake: a possible case of induced seismicity in South Korea. Science
Cao, W., Shi, J.-Q., Durucan, S., Korre, A., 2021. Evaluation of shear slip stress transfer
360 (6392), 1003–1006.
mechanism for induced microseismicity at in Salah CO2 storage site. Int. J.
Guglielmi, Y., Cappa, F., Amitrano, D., 2008. High-definition analysis of fluid-induced
Greenhouse Gas Control 107, 103302.
seismicity related to the mesoscale hydromechanical properties of a fault zone.
Cappa, F., Rutqvist, J., 2011. Impact of CO2 geological sequestration on the nucleation of
Geophys. Res. Lett. 35 (6).
earthquakes. Geophys. Res. Lett. 38 (17).

17
Y. Cheng et al. Earth-Science Reviews 239 (2023) 104369

Hajati, T., Langenbruch, C., Shapiro, S., 2015. A statistical model for seismic hazard Morris, J.P., Detwiler, R.L., Friedmann, S.J., Vorobiev, O.Y., Hao, Y., 2011. The large-
assessment of hydraulic-fracturing-induced seismicity. Geophys. Res. Lett. 42 (24), scale geomechanical and hydrogeological effects of multiple CO2 injection sites on
10601–10606. formation stability. Int.J.Greenhouse Gas Control 5 (1), 69–74.
Hanks, T.C., 1977. Earthquake stress drops, ambient tectonic stresses and stresses that Myer, L.R., Daley, T.M.J.E.P., 2011. Elements of a best practices approach to induced
drive plate motions. In: Stress in the Earth. Springer, pp. 441–458. seismicity in geologic storage. Energy Procedia 4, 3707–3713.
Häring, M.O., Schanz, U., Ladner, F., Dyer, B.C., 2008. Characterisation of the Basel 1 Nicol, A., Carne, R., Gerstenberger, M., Christophersen, A., 2011. Induced seismicity and
enhanced geothermal system. Geothermics 37 (5), 469–495. its implications for CO2 storage risk. Energy Procedia 4, 3699–3706.
Hofmann, H., Zimmermann, G., Farkas, M., Huenges, E., Zang, A., Leonhardt, M., Nordhaus, W.D., 1992. An optimal transition path for controlling greenhouse gases.
Kwiatek, G., Martinez-Garzon, P., Bohnhoff, M., Min, K.-B., 2019. First field Science 258 (5086), 1315–1319.
application of cyclic soft stimulation at the Pohang Enhanced Geothermal System Onuma, T., Ohkawa, S., 2009. Detection of surface deformation related with CO2
site in Korea. Geophys. J. Int. 217 (2), 926–949. injection by DInSAR at In Salah, Algeria. Energy Procedia 1 (1), 2177–2184.
Hortle, A., de Caritat, P., Stalvies, C., Jenkins, C., 2011. Groundwater monitoring at the Orlic, B., Ter Heege, J., Wassing, B., 2011. Assessing the short-term and long-term
Otway project site, Australia. Energy Procedia 4, 5495–5503. integrity of top seals in feasibility studies of geological CO2 storage. In: 45th US Rock
Hovorka, S.D., Meckel, T.A., Trevino, R.H., 2013. Monitoring a large-volume injection at Mechanics/Geomechanics Symposium. OnePetro.
Cranfield, Mississippi—project design and recommendations. Int. J. Greenhouse Gas Ouellet, A., Bérard, T., Desroches, J., Frykman, P., Welsh, P., Minton, J., Pamukcu, Y.,
Control 18, 345–360. Hurter, S., Schmidt-Hattenberger, C., 2011. Reservoir geomechanics for assessing
Hummel, N., Shapiro, S.A., 2016. Back front of seismicity induced by non-linear pore containment in CO2 storage: a case study at Ketzin, Germany. Energy Procedia 4,
pressure diffusion. Geophys. Prospect. 64 (1), 170–191. 3298–3305.
Jia, Y., Fang, Y., Elsworth, D., Wu, W., 2020. Slip velocity dependence of friction- Park, J., Griffiths, L., Dautriat, J., Grande, L., Rodriguez, I.V., Iranpour, K., Bjørnarå, T.I.,
permeability response of shale fractures. Rock Mech. Rock. Eng. 53 (5), 2109–2121. Moreno, H.M., Mondol, N.H., Sauvin, G., 2022. Induced-seismicity geomechanics for
Kaven, J.O., Hickman, S.H., McGarr, A.F., Ellsworth, W.L., 2015. Surface monitoring of controlled CO2 storage in the North Sea (IGCCS). Int. J. Greenhouse Gas Control
microseismicity at the Decatur, Illinois, CO2 sequestration demonstration site. 115, 103614.
Seismol. Res. Lett. 86 (4), 1096–1101. Pawar, R.J., Bromhal, G.S., Carey, J.W., Foxall, W., Korre, A., Ringrose, P.S., Tucker, O.,
Keranen, K.M., Weingarten, M., Abers, G.A., Bekins, B.A., Ge, S., 2014. Sharp increase in Watson, M.N., White, J.A., 2015. Recent advances in risk assessment and risk
Central Oklahoma seismicity since 2008 induced by massive wastewater injection. management of geologic CO2 storage. Int. J. Greenhouse Gas Control 40, 292–311.
Science 345 (6195), 448–451. Payre, X., Maisons, C., Marblé, A., Thibeau, S., 2014. Analysis of the passive seismic
Kim, K.-I., Min, K.-B., Kim, K.-Y., Choi, J.W., Yoon, K.-S., Yoon, W.S., Yoon, B., Lee, T.J., monitoring performance at the Rousse CO2 storage demonstration pilot. Energy
Song, Y., 2018a. Protocol for induced microseismicity in the first enhanced Procedia 63, 4339–4357.
geothermal systems project in Pohang, Korea. Renew. Sustain. Energy Rev. 91, Rathnaweera, T.D., Wu, W., Ji, Y., Gamage, R.P., 2020. Understanding injection-induced
1182–1191. seismicity in enhanced geothermal systems: from the coupled thermo-hydro-
Kim, K., Vilarrasa, V., Makhnenko, R.Y., 2018b. CO2 injection effect on geomechanical mechanical-chemical process to anthropogenic earthquake prediction. Earth Sci.
and flow properties of calcite-rich reservoirs. Fluids 3 (3), 66. Rev. 205, 103182.
Kim, K.H., Ree, J.H., Kim, Y., Kim, S., Kang, S.Y., Seo, W., 2018c. Assessing whether the Richards-Dinger, K., Dieterich, J.H., 2012. RSQSim earthquake simulator. Seismol. Res.
2017 M-w 5.4 Pohang earthquake in South Korea was an induced event. Science 360 Lett. 83 (6), 983–990.
(6392), 1007–1009. Rinaldi, A.P., Jeanne, P., Rutqvist, J., Cappa, F., Guglielmi, Y., 2014a. Effects of fault-
Kim, S., Hosseini, S.A., 2015. Hydro-thermo-mechanical analysis during injection of cold zone architecture on earthquake magnitude and gas leakage related to CO2 injection
fluid into a geologic formation. Int. J. Rock Mech. Min. Sci. 77, 220–236. in a multi-layered sedimentary system. Greenhouse Gases Sci. Technol. 4 (1),
Kisslinger, C., 1976. A review of theories of mechanisms of induced seismicity. Eng. Geol. 99–120.
10 (2–4), 85–98. Rinaldi, A.P., Rutqvist, J., Cappa, F., 2014b. Geomechanical effects on CO2 leakage
Kisslinger, C., Cherry, J., 1970. Excitation of earthquakes by underground explosions, through fault zones during large-scale underground injection. Int. J. Greenhouse Gas
Transactions-American Geophysical Union. Amer Geophysical Union 2000 Florida Control 20, 117–131.
Ave NW, Washington, DC 20009, 353-&amp. Rinaldi, A.P., Vilarrasa, V., Rutqvist, J., Cappa, F., 2015. Fault reactivation during CO2
Kroll, K.A., Buscheck, T.A., White, J.A., Richards-Dinger, K.B., 2020. Testing the efficacy sequestration: effects of well orientation on seismicity and leakage. Greenhouse
of active pressure management as a tool to mitigate induced seismicity. Int. J. Gases Sci. Technol. 5 (5), 645–656.
Greenhouse Gas Control 94, 102894. Rohmer, J., Pluymakers, A., Renard, F., 2016. Mechano-chemical interactions in
Kwiatek, G., Saarno, T., Ader, T., Bluemle, F., Bohnhoff, M., Chendorain, M., Dresen, G., sedimentary rocks in the context of CO2 storage: weak acid, weak effects? Earth Sci.
Heikkinen, P., Kukkonen, I., Leary, P., 2019. Controlling fluid-induced seismicity Rev. 157, 86–110.
during a 6.1-km-deep geothermal stimulation in Finland. Sci.Adv. 5 (5), eaav7224. Rothert, E., Shapiro, S.A., 2007. Statistics of fracture strength and fluid-induced
LaForce, T., Ennis-King, J., Paterson, L., 2015. Semi-analytical temperature and stress microseismicity. J. Geophys. Res. Solid Earth 112 (B4).
profiles for nonisothermal CO2 injection. In: Proceedings of the World Geothermal Rutledge, J., 2011. Southwest Regional Partnership on Carbon Sequestration Phase II.
Congress, Melbourne, Australia. Citeseer, pp. 19–25. New Mexico Institute of Mining and Technology, Socorro, NM (United States).
Le Guen, Y., Huot, M., Loizzo, M., Poupard, O., 2011. Well integrity risk assessment of Rutqvist, J., 2012. The geomechanics of CO2 storage in deep sedimentary formations.
Ketzin injection well (ktzi-201) over a prolonged sequestration period. Energy Geotechn. Geol. Eng. 30 (3), 525–551.
Procedia 4, 4076–4083. Rutqvist, J., Cappa, F., Mazzoldi, A., Rinaldi, A., 2013. Geomechanical modeling of fault
Lee, K.-K., Ellsworth, W.L., Giardini, D., Townend, J., Ge, S., Shimamoto, T., Yeo, I.-W., responses and the potential for notable seismic events during underground CO2
Kang, T.-S., Rhie, J., Sheen, D.-H., 2019. Managing injection-induced seismic risks. injection. Energy Procedia 37, 4774–4784.
Science 364 (6442), 730–732. Rutqvist, J., Liu, H.-H., Vasco, D.W., Pan, L., Kappler, K., Majer, E., 2011. Coupled non-
Linker, M., Dieterich, J.H., 1992. Effects of variable normal stress on rock friction: isothermal, multiphase fluid flow, and geomechanical modeling of ground surface
observations and constitutive equations. J. Geophys. Res. Solid Earth 97 (B4), deformations and potential for induced micro-seismicity at the in Salah CO2 storage
4923–4940. operation. Energy Procedia 4, 3542–3549.
Lucier, A., Zoback, M., Gupta, N., Ramakrishnan, T., 2006. Geomechanical aspects of Rutqvist, J., Rinaldi, A.P., Cappa, F., Jeanne, P., Mazzoldi, A., Urpi, L., Guglielmi, Y.,
CO2 sequestration in a deep saline reservoir in the Ohio River Valley region. Vilarrasa, V., 2016. Fault activation and induced seismicity in geological carbon
Environ. Geosci. 13 (2), 85–103. storage–lessons learned from recent modeling studies. J. Rock Mech. Geotechn. Eng.
Mathieson, A., Midgely, J., Wright, I., Saoula, N., Ringrose, P., 2011. In Salah CO2 8 (6), 789–804.
Storage JIP: CO2 sequestration monitoring and verification technologies applied at Rutqvist, J., Vasco, D.W., Myer, L., 2010. Coupled reservoir-geomechanical analysis of
Krechba, Algeria. Energy Procedia 4, 3596–3603. CO2 injection and ground deformations at in Salah, Algeria. Int. J. Greenhouse Gas
Mazzoldi, A., Rinaldi, A.P., Borgia, A., Rutqvist, J., 2012. Induced seismicity within Control 4 (2), 225–230.
geological carbon sequestration projects: maximum earthquake magnitude and Schneider, S.H., 1989. The greenhouse effect: science and policy. Science 243 (4892),
leakage potential from undetected faults. Int. J. Greenhouse Gas Control 10, 771–781.
434–442. Scholz, C.H., 2019. The Mechanics of Earthquakes and Faulting. Cambridge University
McGarr, A., 2014. Maximum magnitude earthquakes induced by fluid injection. Press.
J. Geophys.Res.Solid Earth 119 (2), 1008–1019. Schwab, D.R., Bidgoli, T.S., Taylor, M.H., 2017. Characterizing the potential for
McGarr, A., Bekins, B., Burkardt, N., Dewey, J., Earle, P., Ellsworth, W., Ge, S., injection-induced fault reactivation through subsurface structural mapping and
Hickman, S., Holland, A., Majer, E., Rubinstein, J., Sheehan, A., 2015. Coping with stress field analysis, Wellington Field, Sumner County, Kansas. J. Geophys. Res. Solid
earthquakes induced by fluid injection. Science 347 (6224), 830–831. Earth 122 (12), 10132–10154.
Mena, B., Wiemer, S., Bachmann, C., 2013. Building robust models to forecast the Segall, P., Lu, S., 2015. Injection-induced seismicity: poroelastic and earthquake
induced seismicity related to geothermal reservoir enhancement. Bull. Seismol. Soc. nucleation effects. J. Geophys.Res.Solid Earth 120 (7), 5082–5103.
Am. 103 (1), 383–393. Shapiro, S., Patzig, R., Rothert, E., Rindschwentner, J., 2003. Triggering of seismicity by
Mendes, R., Costa, A., Sousa, L., Pereira, L., Oliveira, M., 2010. Risks and mitigation pore-pressure perturbations: permeability-related signatures of the phenomenon.
problems in a CO2 injection project for a petroleum onshore field in Brazil. In: 44th Pure Appl. Geophys. 160, 1051–1066.
US Rock Mechanics Symposium and 5th US-Canada Rock Mechanics Symposium. Shapiro, S.A., 2015. Fluid-Induced Seismicity. Cambridge University Press.
OnePetro. Shapiro, S.A., Dinske, C., Kummerow, J., 2007. Probability of a given-magnitude
Mit, 2022. Carbon Capture and Sequestration Technologies. Massachusetts Institute of earthquake induced by a fluid injection. Geophys. Res. Lett. 34 (22).
Technology. Shapiro, S.A., Dinske, C., Langenbruch, C., Wenzel, F., 2010. Seismogenic index and
magnitude probability of earthquakes induced during reservoir fluid stimulations.
Lead. Edge 29 (3), 304–309.

18
Y. Cheng et al. Earth-Science Reviews 239 (2023) 104369

Shapiro, S.A., Krüger, O.S., Dinske, C., Langenbruch, C., 2011. Magnitudes of induced Vilarrasa, V., Carrera, J., Olivella, S., Rutqvist, J., Laloui, L., 2019. Induced seismicity in
earthquakes and geometric scales of fluid-stimulated rock volumes. Geophysics 76 geologic carbon storage. Solid Earth 10 (3), 871–892.
(6), WC55-WC63. Vilarrasa, V., Laloui, L., 2015. Potential fracture propagation into the caprock induced by
Singh, A.-K., Böttcher, N., Wang, W., Park, C.-H., Görke, U.-J., Kolditz, O., 2011. Non- cold CO2 injection in normal faulting stress regimes. Geomech. Energy Environ. 2,
isothermal effects on two-phase flow in porous medium: CO2 disposal into a saline 22–31.
aquifer. Energy Procedia 4, 3889–3895. Vilarrasa, V., Laloui, L., 2016. Impacts of thermally induced stresses on fracture stability
Soma, N., Rutledge, J.T., 2013. Relocation of microseismicity using reflected waves from during geological storage of CO2. Energy Procedia 86, 411–419.
single-well, three-component array observations: application to CO2 injection at the Vilarrasa, V., Makhnenko, R., Gheibi, S., 2016. Geomechanical analysis of the influence
Aneth oil field. Int. J. Greenhouse Gas Control 19, 74–91. of CO2 injection location on fault stability. J. Rock Mech. Geotechn. Eng. 8 (6),
Stirling, M., Litchfield, N., Gerstenberger, M., Clark, D., Bradley, B., Beavan, J., 805–818.
McVerry, G., Van Dissen, R., Nicol, A., Wallace, L., 2011. Preliminary probabilistic Vilarrasa, V., Olivella, S., Carrera, J., Rutqvist, J., 2014. Long term impacts of cold CO2
seismic hazard analysis of the CO2CRC Otway Project Site, Victoria, Australia. Bull. injection on the caprock integrity. Int. J. Greenhouse Gas Control 24, 1–13.
Seismol. Soc. Am. 101 (6), 2726–2736. Vilarrasa, V., Rutqvist, J., 2017. Thermal effects on geologic carbon storage. Earth Sci.
Stork, A., Nixon, C., Hawkes, C., Birnie, C., White, D., Schmitt, D., Roberts, B., 2018. Is Rev. 165, 245–256.
CO2 injection at Aquistore aseismic? A combined seismological and geomechanical Vilarrasa, V., Rutqvist, J., Rinaldi, A.P., 2015. Thermal and capillary effects on the
study of early injection operations. Int. J. Greenhouse Gas Control 75, 107–124. caprock mechanical stability at in Salah,Algeria. Greenhouse Gases: Science
Talwani, P., Acree, S., 1985. Pore pressure diffusion and the mechanism of reservoir- Technology 5 (4), 449–461.
induced seismicity. In: Earthquake Prediction. Springer, pp. 947–965. Vilarrasa, V., Silva, O., Carrera, J., Olivella, S., 2013. Liquid CO2 injection for geological
Torp, T.A., Gale, J., 2004. Demonstrating storage of CO2 in geological reservoirs: the storage in deep saline aquifers. Int. J. Greenhouse Gas Control 14, 84–96.
Sleipner and SACS projects. Energy 29 (9–10), 1361–1369. Wenzel, F., 2015. Induced seismicity using dieterich’s rate and state theory and
Urpi, L., Rinaldi, A.P., Rutqvist, J., Cappa, F., Spiers, C.J., 2016. Dynamic simulation of comparison to the critical pressure theory. Energy Procedia 76, 282–290.
CO2-injection-induced fault rupture with slip-rate dependent friction coefficient. Wenzel, F., 2017. Fluid-induced seismicity: comparison of rate-and state-and critical
Geomech. Energy Environ. 7, 47–65. pressure theory. Geotherm. Energy 5 (1), 1–16.
Van der Elst, N.J., Page, M.T., Weiser, D.A., Goebel, T.H., Hosseini, S.M., 2016. Induced White, J.A., Foxall, W., 2016. Assessing induced seismicity risk at CO2 storage projects:
earthquake magnitudes are as large as (statistically) expected. J. Geophys. Res. Solid recent progress and remaining challenges. Int. J. Greenhouse Gas Control 49,
Earth 121 (6), 4575–4590. 413–424.
Verdon, J., Kendall, J.-M., White, D., Angus, D., 2011. Linking microseismic event Whittaker, S., Rostron, B., Hawkes, C., Gardner, C., White, D., Johnson, J.,
observations with geomechanical models to minimise the risks of storing CO2 in Chalaturnyk, R., Seeburger, D., 2011. A decade of CO2 injection into depleting oil
geological formations. Earth Planet.Sci.Lett. 305 (1–2), 143–152. fields: monitoring and research activities of the IEA GHG Weyburn-Midale CO2
Verdon, J.P., Kendall, J.-M., White, D.J., Angus, D.A., Fisher, Q.J., Urbancic, T., 2010. monitoring and Storage Project. Energy Procedia 4, 6069–6076.
Passive seismic monitoring of carbon dioxide storage at Weyburn. Lead. Edge 29 (2), Wiese, B., Nimtz, M., Klatt, M., Kühn, M., 2010. Sensitivities of injection rates for single
200–206. well CO2 injection into saline aquifers. Geochemistry 70, 165–172.
Verdon, J.P., Stork, A.L., 2016. Carbon capture and storage, geomechanics and induced Würdemann, H., Möller, F., Kühn, M., Heidug, W., Christensen, N.P., Borm, G.,
seismic activity. J. Rock Mech. Geotechn. Eng. 8 (6), 928–935. Schilling, F.R., Group, C.S., 2010. CO2SINK—From site characterisation and risk
Verdon, J.P., Stork, A.L., Bissell, R.C., Bond, C.E., Werner, M., 2015. Simulation of assessment to monitoring and verification: one year of operational experience with
seismic events induced by CO2 injection at in Salah, Algeria. Earth Planet. Sci. Lett. the field laboratory for CO2 storage at Ketzin, Germany. International Journal of
426, 118–129. Greenhouse Gas Control 4 (6), 938–951.
Vialle, S., Vanorio, T., 2011. Laboratory measurements of elastic properties of carbonate Yadav, A., Gahalaut, K., Purnachandra Rao, N., 2017. 3-D modeling of pore pressure
rocks during injection of reactive CO2-saturated water. Geophys. Res. Lett. 38 (1). diffusion beneath Koyna and Warna reservoirs,Western India. Pure Appl. Geophys.
Vidal-Gilbert, S., Nauroy, J.-F., Brosse, E., 2009. 3D geomechanical modelling for CO2 174 (5), 2121–2132.
geologic storage in the Dogger carbonates of the Paris Basin. Int. J. Greenhouse Gas Zang, A., Yoon, J.S., Stephansson, O., Heidbach, O., 2013. Fatigue hydraulic fracturing
Control 3 (3), 288–299. by cyclic reservoir treatment enhances permeability and reduces induced seismicity.
Vidal-Gilbert, S., Tenthorey, E., Dewhurst, D., Ennis-King, J., Van Ruth, P., Hillis, R., Geophys. J. Int. 195 (2), 1282–1287.
2010. Geomechanical analysis of the Naylor Field, Otway Basin, Australia: Zhang, W., Li, Y., Xu, T., Cheng, H., Zheng, Y., Xiong, P., 2009. Long-term variations of
implications for CO2 injection and storage. Int. J. Greenhouse Gas Control 4 (5), CO2 trapped in different mechanisms in deep saline formations: a case study of the
827–839. Songliao Basin, China. Int. J. Greenhouse Gas Control 3 (2), 161–180.
Vilarrasa, V., 2016. The role of the stress regime on microseismicity induced by Zoback, M.D., Gorelick, S.M., 2012. Earthquake triggering and large-scale geologic
overpressure and cooling in geologic carbon storage. Geofluids 16 (5), 941–953. storage of carbon dioxide. Proc. Natl. Acad. Sci. 109 (26), 10164–10168.

19

You might also like