You are on page 1of 17

SPE-195807-MS

Numerical Study of Injection-Induced Seismicity Using A FEM-BEM


Coupling Approach

Masato Aoki, Kentaro Kimura, Yoichiro Iijima, and Kenji Furui, Waseda University

Copyright 2019, Society of Petroleum Engineers

This paper was prepared for presentation at the SPE Annual Technical Conference and Exhibition held in Calgary, Alberta, Canada, 30 Sep - 2 October 2019.

This paper was selected for presentation by an SPE program committee following review of information contained in an abstract submitted by the author(s). Contents
of the paper have not been reviewed by the Society of Petroleum Engineers and are subject to correction by the author(s). The material does not necessarily reflect
any position of the Society of Petroleum Engineers, its officers, or members. Electronic reproduction, distribution, or storage of any part of this paper without the written
consent of the Society of Petroleum Engineers is prohibited. Permission to reproduce in print is restricted to an abstract of not more than 300 words; illustrations may
not be copied. The abstract must contain conspicuous acknowledgment of SPE copyright.

Abstract
Induced seismicity caused by underground fluid injection occurs because of pore pressure changes that lead
to stress changes in the reservoir and the surrounding formations. Despite that noticeable seismic events
from fluid injection is very rare, proper assessment of possible seismic events is important. The objective
of this study is to develop numerical models that simulate stress changes, fault slips, and ground floor
movements induced by underground fluids injection.
The numerical analysis process presented in this work consists of three steps. First, stress changes around
the reservoir due to fluid injection are analyzed using a FEM-BEM (Finite Element Method - Boundary
Element Method) coupled model. Secondly, the stability of faults located near the reservoir is evaluated
using the displacement discontinuity method. Thirdly, elastic waves caused by the fault slip is simulated
using a FEM model where seismic response on the surface are calculated. A field case study is also presented
to demonstrate the applicability of the numerical model developed in this work.
The numerical analysis results indicate that stress concentration occurs around a boundary between the
basement and sandstone beneath the reservoir. This affects the stability of existing faults in this region. As
a result, when the fault is slipped, seismicity may be triggered. It is assumed that the slip is caused by stress
changes in the faulted region as well as a pore pressure change in the fault, which is caused by volumetric
strain changes of the fluid in the fault. A field case study based on wastewater injection in the Southwestern
region of the United States where injection induced seismicity events have been recently reported, is also
performed in this work. In this case study, the variation of rock strength is considered one of important
factors in induced seismicity events.
The novelty of our model is the ability to quantitatively assess the risk of induced seismicity due
to wastewater injection, which can be also applied to other applications such as CCS and underground
gas storages. Moreover, conducting risk assessment by these numerical models can improve safety of
underground fluid injection operations.

Introduction
It has been widely known that earthquakes can be induced by impoundment of water reservoirs, surface
and underground mining, withdrawal of water, oil, and gas from the subsurface, and injection of fluids
2 SPE-195807-MS

into underground formations. As the application of hydraulic fracturing to tight sand and shale formations
has grown rapidly, especially in the North America, injection-induced earthquakes have become a focus of
discussion in the industry.
Foulger et al. (2018) reviewed and analyzed numerous seismicity data obtained from a human-
induced earthquake database, HiQuake (Wilson, 2017) including various industrial projects such as the
impoundment of water reservoirs, quarrying, extraction of groundwater, coal, minerals, gas oil and
geothermal fluids, etc. They argued that the minimum amount of stress changes to induce earthquake is
very difficult to determine since earthquakes occur naturally without any human intervention at all. In some
cases, it is reported that induced stress changes of only less than 0.1 MPa may have triggered earthquakes
with magnitudes greater than 4.
The 2016 M 5.8 Pawnee earthquake in the Oklahoma is the largest seismicity by injection water (Hincks
et al., 2018). In Oklahoma, seismicity had almost not occurred before injection fluid, but the number of
seismic events had been increasing after the beginning of injecting wastewater. The induced seismicity
caused damage to the residents and property. USGS reported that the seismicity in Oklahoma were caused
by wastewater injection, because the events occurred at shallow depth unlike in the case of plate tectonics.
The number of seismic events has been decreasing since the regulation of injection begun, that is evidence
to show the correlation between earthquake frequency and volume of wastewater injected into underground
formations.
In Colorado, since injecting wastewater from oil and gas production in 1994, over 90 seismic events
greater than M3.0 have occurred, M5.3 at most (Matthew et al., 2009). The earthquakes nearly all occurred
within 5- 10 km of injection wells. Therefore, the fact indicates the connection between seismicity
and wastewater injection. However, the Colorado case differs from classic one. Although the seismicity
happened spatially near injection wells, the majority occurred at depth 0.5-7 km below injection interval.
In the classic case, the seismic event occurs at injection depth, because the pressure change is the largest
near wastewater injection area.
Except oilfield examples mentioned above, it is also reported that earthquakes were induced by
geothermal production and water impoundment behind dams. The induced seismicity events caused by
geothermal operations have been reported in the south Korea (Kim et al., 2018) and at the Geysers, USA
(Foulger et al., 2018) while seismicity events related to dams were observed at the Koyna in India (Gupta,
1985) and Hoover in USA (Meade, 1991).
As widely agreed, there is presently no reliable method to predict earthquakes. Current approaches to
reducing hazard comprise long-term forecasting based on the history of past earthquakes. This approach
assumes that patterns of seismicity persist in local areas and not applicable to other areas. One of the biggest
challenges in earthquake prediction is that it is very difficult to determine how much stress changes are
required to induce fault slips. On the other hand, it is very important to properly assess the risk of induced
seismicity by understanding pore pressure and stress changes caused by fluids injection and extraction in
underground formations.
The objective of our study is to develop a numerical analysis process that simulates stress changes, fault
slips, and ground floor movements induced by underground fluids injection.

Development of Numerical Models for Induced Seismicity Analysis


The numerical analysis process developed in this work consists of three steps. At the first step, stress changes
within and around the reservoir due to fluid injection are analyzed using a 2D FEM-BEM coupled model.
In this model, the reservoir is discretized by finite elements while the boundary of the reservoir and the
surface are discretized by boundary elements. Secondly, the stability of faults located near the reservoir is
evaluated using the displacement discontinuity method with the joint element model. Thirdly, elastic waves
SPE-195807-MS 3

caused by the fault slip is simulated using a dynamic elastic FEM model where seismic response on the
surface are calculated. In the following section, each model is briefly explained.

Stress Analysis Using a Coupled FEM-BEM Model


Figure 1 illustrates a conceptual mesh geometry employed for the coupled FEM-BEM model. In this work,
plane strain conditions are assumed. The reservoir section where pore pressure changes are anticipated
are modeled by finite elements denoted by ΩFE (FE region). In the FEM model, the pore pressure term is
included in the stress-strain relationships:

(1)

Figure 1—Discretization of the domain by finite and boundary elements

where E and ν are Young's modulus and Poisson's ratio of rock, Em and νm are Young's modulus and Poisson's
ratio of matrix, εij is strain, is the effective stress, p is pore pressure, and δij is the Kronecker delta. The
free surface and the reservoir boundary are discretized by boundary elements simulating stress and strain
changes in the formations surrounding the reservoir denoted by ΩBE (BE region). It should be noted that, to
improve the computational efficiency, infinite elements are used for discretizing the free surface.
The boundary between the FE and BE regions, ΛI, share nodes where the boundary nodal tractions are
converted into equivalent nodal point forces for coupling the FEM and BEM regions.
For the BE region, the boundary integral equation can be expressed by the following matrix form
(2)
where, U is displacement vector, T is traction vector. H and G are boundary element matrices obtained from
the boundary integral equation. For the FE region, the stiffness equations can be written in the matrix form:
(3)
where, K is stiffness matrix and F is nodal force vector. Equations 2 and 3 can be coupled using the
equilibrium and compatibility conditions at the interface:
(4)
(5)
Multiplying G-1 in the both sides of Eq. 2 to give
(6)
The traction vector can be transformed into nodal forces by multiplying with the interpolation function
N, which are used to obtain displacements of the BE region.
4 SPE-195807-MS

(7)
or
(8)
where
(9)
(10)
Finally, the system of simultaneous linear equations is obtained as follows.
(11)
where

(12)

(13)

For all region, displacements and nodal forces are obtained by solving Eq. 11.

Fault Slip Analysis Using Displacement Discontinuity Method


The fault slip analysis model is based on the joint analysis method proposed by Crouch and Starfield (1974).
Following their work, a fault is considered as a long, thin crack with a compressible filling. A segment of a
joint can then be modeled as an elemental displacement discontinuity whose opposite surfaces are connected
by a material with the normal and shear stiffnesses, Kn and Ks chosen to be representative of the properties of
the joint filling material. The joint filling material is inelastic. According to the Mohr-Coulomb condition,
(14)
where c and ϕ are the cohesion and angle of friction of the join-filling material of element i. Assuming that
the fault is discretized into N displacement discontinuity elements, the governing equations for the normal
and shear deformations at the i-th Mohr-Coulomb element for loading increment k are given by

(15)

(16)

where Ans, Ann, Ass, and Asn are the boundary influence coefficients and Xs and Xn are displacement
discontinuities for shear and normal directions, respectively. Also Pfault is pore pressure in the fault. If the

element is yielding the total shear stress must be equal to the yield stress , and Eq. 16 becomes

(17)
SPE-195807-MS 5

Seismic Response Analysis Using Split Node Method


In this study, seismic waves induced by fault slippage is calculated by a 3D FEM model for the vibration
analysis of a three-dimensional elastic solid. The fault slip is modeled using the split node technique
proposed by Melosh and Raefsky (1981). In the split node method, a fault displacement at a single node
point in the FEM mesh shared between two elements. This method does not require modifications of the
stiffness matrix, and it allows any node of the grid to be split without adding new degrees of freedom.
In a conventional FEM model, generally local displacement is equal at shared point between different
elements. However, in the split node method, local displacement is not equal.
Figure 2 shows the application of the split node method for a 1D problem. When dividing node 2 in the
global coordinate, the displacement represented in element 1 and 2 are shown to be
(18)
(19)

Figure 2—One dimensional FEM of Node Split Method

where the superscript i in Uij represents the element number and the subscript j is the local node number.
Also represents the mean value of displacement, which is continuous between the two elements:
(20)
On the other hand, is discontinuous between the elements. The split node term is described by
(21)
In element 1, the local stiffness equation can be written by

(22)

The above equation can be also written as

(23)

Similarly, the stiffness equation can be written for element 2. Assembling these equations into the global
system of equations to give,
6 SPE-195807-MS

(24)

Note that splitting induces fictitious forces on the nodes adjacent to the split node, as well as the split
node itself. In general, such forces appear at every node of each element which has a split node. The load
vector is not modified at the nodes of other elements. In this work, displacement discontinuities calculated
by the fault slip analysis model are used for dislocation ΔU in the seismic response model.

Integration Method of Numerical Models


Figure 3 shows the workflow of the injection induced seismicity analysis method proposed in this work.

Figure 3—Integrated Workflow of Injection-Induced Seismicity Analysis Method.

Step 1) Injection-induced stress analysis – In this step, the dimensions of reservoir and overburden,
underburden, and sideburden formations are defined creating a mesh used for the stress analysis. Pore
pressure changes in the reservoir, which can be estimated from any commercial reservoir simulation
software are used as input in the model. Stress changes within the reservoir and surrounding formations are
calculated by solving Eq. 11. It should be noted that the model does not contain any fault at this stage.
Step 2) Fault slip analysis – Stress changes expected in a particular fault location can be obtained from
the stress analysis performed in Step 1. These stress changes are used as input data for the displacement
discontinuity model (DDM). An assumed fault geometry, preferably defined from seismic data and a
geological analysis, is created in the DDM model. Normal and tangential stress changes are applied to the
fault elements. Having fault properties assigned, a fault slip analysis is performed estimating shear and
normal displacements along the fault.
Step 3) The fault movements estimated in Step 2 is applied to the split nodes (i.e., fault location) at a
certain fault slip velocity in the dynamic elastic FEM model. The model calculates dynamic displacements
and synthetic accelerations at the surface, which can be used for accessing risk of particular seismic events
experienced during underground fluids injection.

Case Study
In this section, we present a field case study of induced seismicity in the Raton Basin in southern Colorado
and northern New Mexico to demonstrate the applicability of the numerical model developed in this work.
SPE-195807-MS 7

Some geological data was obtained from the research paper (Matthew et al., 2015) and the USGS Model
Viewer (Matthew et al., 2009).

Stress Analysis Result


The model is shown in Fig.4, which have crystalline basement rock layer located 3.0 km depth beneath
reservoir. The injection wells interval is located 1.5 km depth, and the formation thickness is 200 m. The
pore pressure increases are assumed to be 3.0 MPa. In this field, the distance between wastewater injection
reservoir and basement rock is 1.5 km. The rock properties are summarized in Table 1. In this work, it is
assumed that the Young's modulus of the basement rock is about 20 times larger than one of the reservoir.

Figure 4—Distribution of shear stress changes around the injection reservoir.

Table 1—Rock Properties for Stress Analysis Result

Figure 3 shows the distribution of shear stress (τxy) changes denoted by Sxy in the figure. These results
indicate that the shear stress changes near and at the boundary between the sandstone and the crystalline
basement becomes larger than the surrounding area indicating that stress concentration occurs in the area
where a large rock strength contrast exists. Therefore, if the fault exists at the boundary, stress change by
injection water affects the fault stability. In addition, it may be possible that the stress concentration produces
shear fractures by brittle failure. (Mogi, 1974)

Fault Slip Analysis Result


Several east-vergin thrust faults deform and overturn a lot of the sedimentary of the deep Raton Basin. It is
reported that the fault zone is 15 km long and 2 km wide at the top of crystalline basement (MATTHEW
et al., 2009).
8 SPE-195807-MS

In this analysis, the volume of fault slippage by pore pressure change is also calculated. Analysis
conditions are summarized in Table 2. Shear modulus is obtained by Eq. 25. (Scholz et al., 2002)

(25)

where L is fault length.

Table 2—Rock Properties for Fault Slip Analysis

Figure 6 shows the distribution of shear displacement in the fault. The result shows that shear
displacements 1m occur at the part of fault 1 km. This phenomenon is typically. In most of fault slippage,
fault slips partly. This result is equivalent to magnitude 5, which is near the magnitude of seismicity occurred
in the Colorado case (Fig. 5).

Figure 5—Distribution of earthquake hypocenters in the Raton Basin in 2011 (Matthew et al., 2009).
SPE-195807-MS 9

Figure 6—shear displacement of fault

Based on the volumetric stress changes estimated from the stress analysis (Step 1), the pore pressure
increase in the fault is estimated to be approximately 0.5 MPa when stress change occurs in the formation.
Below, we present a sensitivity analysis of fault slippage considering various pore pressure changes in fault.
Analysis conditions are summarized in Table 3.

Table 3—Rock Properties for Fault Slip Analysis

Fig. 7 shows that the share displacement of the fault increases as the pore pressure change in the fault
increases.
10 SPE-195807-MS

Figure 7—shear displacement of fault by pore pressure change in fault

Seismic Response Analysis


Figure 8 illustrates a 3D FEM mesh employed for the seismic response analysis in this work. Analysis
conditions are summarized in Table 3. The fault exists on the red line and is located at the depth of 3-4 km.
The friction angle is set as 63 degree. The yellow point is observation point at the surface where we measure
dynamic displacements and acceleration. The fault slipping velocity is assumed to be 5 m/sec in this study.
We set the location of hypocenter at the boundary between basement rock and sandstone where the stress
concentration occurred. The fault slippage is 0.5 m estimated by the fault slip analysis. To avoid boundary
effects, the width of the field model is extended to 9,000 m.

Figure 8—Model overview for Seismic Response Analysis


SPE-195807-MS 11

Table 4—Rock Properties for Seismic Response Analysis

Figure 9 shows displacement (vibration) while Fig. 10 shows synthetic acceleration at the observation
point. The simulation results indicate that the dynamic displacements and acceleration measured at the
ground surface are relatively large, that is, people near the observation point could feel the ground moving
and hanging objects may move slightly.

Figure 9—seismic wave at the observation point


12 SPE-195807-MS

Figure 10—synthetic acceleration at the observation point

Seismic intensity map


In this section, the Japanese seven-stage seismic intensity scale is made into a map in the case of each
M4, M5, and M6 when injecting fluid. The seismic intensity scale is established by the Japan Metrological
Agency (JMA) seismic intensity scale (Table 6 (JMA. 2018.)). In this case, a fault exists 2km depth. The
depth of the fault and slipping velocity are the same values in all cases. The analytical conditions of each
case are summarized in Table 5.

Figure 11—Model overview


SPE-195807-MS 13

Table 5—Rock Properties and analysis conditions

Table 6—Human perception and reaction, indoor situation, outdoor situation (JMA. 2018).

Fig. 12 presents that seismicity that residents feel a quake occurs within small range in the case of M4.
However, Fig. 13-15 present that as the magnitude is larger, the range of over seismic intensity 2 is wider.
14 SPE-195807-MS

The results also present that seismicity over seismic intensity 4 occur widely if the magnitude is large. There
is possibility that seismicity of seismic intensity 5 occurs if the magnitude is approximately 6, the induced
earthquake may destroy concrete block walls.

Figure 12—The range that event over seismic intensity 2 occurs in the case of M4, M5, and M6

Figure 13—The range that event over seismic intensity 3 occurs in the case of M5and M6

Figure 14—The range that event over seismic intensity 4 occurs in the case of M5and M6
SPE-195807-MS 15

Figure 15—The range that event over seismic intensity 5 occurs in the case of M6

Conclusion
The following conclusions can be drawn from our study:

• In this study, the three-step analysis method, that consists of the stress analysis using the coupled
FEM-BEM model, the fault slip analysis using the displacement discontinuity method, and
the seismic response analysis using the split-node method, is presented to investigate induced
seismicity events experienced during fluids injection in underground formations.
• The coupled FEM-BEM stress analysis model indicates that relatively large stress changes are
expected near the interface between the sandstone layer and crystalline basement where a relatively
large rock strength contrast is anticipated. The analysis result may explain the occurrence of major
seismic events in the upper part of the crystalline basement often reported in Oklahoma and New
Mexico-Colorado areas during wastewater injection.
• The fault stability analysis model shows that the stability of the fault may be very sensitive to the
pore pressure in the fault. If the fluid (water) in the fault is trapped, induced stress changes can
cause some pore pressure increases that may increase likelihood of fault slippage to some extent.
• Using the numerical models developed in this work, induced seismicity events observed in the
Raton Basin is analyzed. The analysis result shows that the 3.0 MPa pore pressure changes cause
shear stress changes of 0.6 MPa at the top of the crystalline basement leading a relatively large
seismicity event equivalent to M5. Although the magnitude of actual seismic events depends
greatly on assumptions of fault properties and pore pressures within the fault, these analysis results
may qualitatively explain the process of the induced seismicity in the Raton Basin case.
• The three-step analysis method and numerical models developed in this study can be used to assess
the risk of induced seismicity due to wastewater injection and other injection activities such as CCS
(Carbon Capture and Storage), underground gas storage, water injection in geothermal production.

Acknowledgement
A part of the present work was performed as a part of activities of Research Institute of Sustainable Future
Society, Waseda Research Institute for Science and Engineering, Waseda University.

Nomenclature
εij =strain
σije =elemnt strain
Em =young's modulus of grain
16 SPE-195807-MS

νm =poisson's ratio of grain


p =pore pressure
δij =kronecker delta
pfault =pore pressure in fault
K =stiffness matrix
U =displacement vector
F = nodal force vector
H =coefficient matrix
G =coefficient matrix
T =traction vector
K BE =stiffness matrix of BE region
K(FE) =stiffness matrix of FE region
UIBE =displacement vector of BE region on interface
UIFE =displacement vector of FE region on interface
FIBE =nodal force vector of BE region on interface
FIFE =nodal force vector of FE region on interface
N =shape fuction matrix
Ks =share modulus
Kn =normal modulus
M =the number of fictious stress element
N =the number of element
E =young's modulus
ν =poisson's ratio
A =the boundary influence coefficient
X =unknown quantities
σs =share stress
σn =normal stress
c =cohesion
ϕ =angle of friction of joint filling material
ΔU =dislocation

References
Beer, G., Smith, I. M., and Duenser, C. 2008. The Boundary Element Method with Programing. Wien: Springer-Verlag/
Wien.
Weingarten, M. B. 2009. On the interaction between fluids and earthquakes in both natural and induced seismicity. BS
thesis, University of Wisconsin-Madison, Madison, Wisconsin (2009).
Smith, M., Griffiths, D. V., and Margetts, L. 2014. Programming the Finite Element Method, fifth edition. West Sussex:
John Wiley & Sons Ltd.
Merosh, H. J. and Raefsky, A. 1981. A Simple and Efficient Method for Introducting Faults into Finite Element
Computation. Bulletin of the Seismological Society of America 71(5): 1391–1400.
Freund, B. and Barnett, D. M. 1976. A Two-Dimensional Analysis of Surface Deformation Due to Dip-Slip Faulting.
Bulletin of the Seismological Society of America 66(3): 667–675.
Crouch, S. L. and Starfield, A. M. 1983. BOUNDARY ELEMENT METHODS IN SOLID MECHANICS. London: George
Allen & Unwin Ltd.
Iida, K. 1965. Earthquake magnitude, earthquake fault, and source dimensions. The Journal of earth sciences, Nagoya
University 13(2): 115–132.
Brebbia, C. A. and Dominguez, J. 1996. Boundary Elements: An Introductory Course, second edition. Southampton: WIT
Press/Computational Mechanics Publications.
Brebbia, C. A., Telles, L.C.F., WROBEL, L.C. 1984. Boudary Element Techniques. Heidelberg: Springer-Verlag Berlin.
SPE-195807-MS 17

Scholz, C. H. 2002. The Mechanics of Earthquakes and Faulting, second edition. Cambridge: The Press Syndicate of
The University of Cambidge.
Mogi, K. 1974. Earthquakes as fractures in the Earth. Bulletin of the Earthquake Research Institute, University of Tokyo
1(A): 559–568.
Walker, R. L. and Aminzadeh, F. 2016. Dynamic Considerations for Induced Seismicity. Presented at the SPE Western
Regional Meeting, Anchorage, Alaska, 23-26 May. SPE-180365-MS. https://doi.org/10.2118/180365-MS
Hincks, T., Aspinall, W., Cooke, R. et al. 2018. Oklahoma's induced seismicity strongly linked to wastewater injection
depth. Science 359(6381): 1251–1255.
Kim, K., Ree, J., Kim, Y. et al. 2018. Assessing whether the 2017 Mw 5.4 Pohang earthquake in South Korea was an
induced event. Science 360(6392): 1007–1009.
Foulger, G. R., Wilson, M. P., Gluyas, J. G. et al. 2018. Global review of human-induced earthquakes. Earth-Science
Reviews 178: 438–514.
Meade, R. B. 1991. Reservoirs and earthquakes. Engineering Geology 30(3-4): 245–262
Gupta, H. K. 1985. The Present Status of Reservoir Induced Sesimicity Investigation with Special Emphasis on Koyna
Earthquakes. Tectonophysics 118(3-4): 257–279.
U.S.G.S. 2002. Investigation of an Earthquake Swarm near Trinidad, Colorado, August-October 2001. Open-File Report,
Contact No. 2002-73, US Geological Survey, Colorado (2002).
JMA. 2018. Japan Meteorological Agency. Tables explaining the JMA Seismic Intensity Scale, 30 January 2018, http://
www.jma.go.jp/jma/en/Activities/inttable.html (accessed 10 June 2019).
Wilson, M. P., Foulger, G. R., Gluyas, J. G. et al. 2017. HiQuake: the Human-Induced Earthquake Database. Seismological
Research Letters. 88(6):1560–1565.
Melosh, H. J. and Raefsky, A. 1981. A Simple and Efficient Method for Introducing Faults into Finite Element
Computations. Bulletin of the Seismological Society of America. 71(5): 1391–1400.

You might also like