Fluid Induced Seismicity

FLUID-INDUCED SEISMICITY
The characterization of fluid-transport properties of rocks is one of the most impor-

tant, yet difficult, challenges of reservoir geophysics, but is essential for optimal
development of hydrocarbon and geothermal reservoirs. Production of shale oil,
shale gas, heavy oil and geothermal energy, as well as carbon-dioxide seques-
tration, are relatively recent developments where borehole fluid injection is often
employed to enhance fluid mobility. Unlike active seismic methods, which present
fundamental difficulties for estimating the permeability of rocks, microseismic-
ity induced by fluid injection in boreholes provides the potential to characterize
physical processes related to fluid mobility and hydraulic-fracture growth in rocks.
This book provides a quantitative introduction to the underlying physics,
application, interpretation, and hazard aspects of fluid-induced seismicity with a
particular focus on its spatio-temporal dynamics. It presents many real-data exam-
ples of microseismic monitoring of hydraulic fracturing at hydrocarbon fields and
of stimulations of enhanced geothermal systems. The author also covers introduc-
tory aspects of linear elasticity and poroelasticity theory, as well as elements of
seismic rock physics and of the mechanics of earthquakes, enabling readers to
develop a comprehensive understanding of the field. Fluid-Induced Seismicity is
a valuable reference for researchers and graduate students working in the fields
of geophysics, geology, geomechanics and petrophysics, and a practical guide for
petroleum geoscientists and engineers working in the energy industry.
SERGE A. SHAPIRO is Professor of Geophysics at the Freie Universität Berlin,

and since 2004, Director of the PHASE (PHysics and Application of Seismic
Emission) university consortium project. From 2001 to 2008 he was one of the
coordinators of the German Continental Deep Drilling Program. His research inter-
ests include seismogenic processes, wave phenomena, exploration seismology and
rock physics. He received the SEG Virgil Kauffman Gold Medal in 2013 for his
work on fluid-induced seismicity and rock physics, and in 2004 was elected a
Fellow of the Institute of Physics.
FLUID-INDUCED SEISMICITY
S e rg e A . S h a p i r o
Freie Universität Berlin
University Printing House, Cambridge CB2 8BS, United Kingdom
Cambridge University Press is part of the University of Cambridge.

It furthers the University’s mission by disseminating knowledge in the pursuit of
education, learning and research at the highest international levels of excellence.
www.cambridge.org
Information on this title: www.cambridge.org/9780521884570

c Serge A. Shapiro 2015
This publication is in copyright. Subject to statutory exception
and to the provisions of relevant collective licensing agreements,
no reproduction of any part may take place without the written
permission of Cambridge University Press.
First published 2015
Printing in the United Kingdom by TJ International Ltd., Padstow, Cornwall
A catalog record for this publication is available from the British Library
Library of Congress Cataloging in Publication Data
Shapiro, S. A.
Fluid-induced seismicity / Serge A. Shapiro, Earth Science Department, Freie Universität Berlin.
pages cm
Includes bibliographical references and index.
ISBN 978-0-521-88457-0
1. Rock mechanics. 2. Hydraulic fracturing. 3. Induced seismicity. 4. Reservoir-triggered
seismicity. 5. Oil field flooding. I. Title.
QE431.6.M4S53 2015
551.22–dc23
2014043958
ISBN 978-0-521-88457-0 Hardback
Cambridge University Press has no responsibility for the persistence or accuracy
of URLs for external or third-party internet websites referred to in this publication,
and does not guarantee that any content on such websites is, or will remain,
accurate or appropriate.
Contents
Preface page ix
Acknowledgments xii
1 Elasticity, seismic events and microseismic monitoring 1
1.1 Linear elasticity and seismic waves 1
1.1.1 Strain 2
1.1.2 Stress 2
1.1.3 Stress–strain relations 4
1.1.4 Elastic moduli 6
1.1.5 Dynamic equations and elastic waves 10
1.1.6 Point sources of elastic waves 12
1.1.7 Static equilibrium 13
1.2 Geomechanics of seismic events 14
1.2.1 Faults and principal stresses 15
1.2.2 Friction coefficient 20
1.2.3 Growth of finite cracks: a sufficient condition 22
1.2.4 Necessary conditions of crack growth and some results
of fracture mechanics 23
1.2.5 Earthquake motions on faults 26
1.3 Elastic wavefields radiated by earthquakes 30
1.4 Introduction to microseismic monitoring 37
1.4.1 Detection of seismic events 37
1.4.2 Seismic multiplets 38
1.4.3 Location of seismic events 42
1.4.4 Microseismic reflection imaging 46
2 Fundamentals of poroelasticity 48
2.1 Linear stress–strain relations in poroelastic materials 49
2.2 Linear stress–strain relations in fluid-saturated materials 55
v
vi Contents
2.2.1 Independent confining stress and pore pressure 55

2.2.2 Linear stress–strain relations in undrained media 59
2.3 Fluid flow and dynamic poroelasticity 62
2.4 Poroelastic wavefields 65
2.4.1 Dispersion relations for poroelastic wavefields 67
2.4.2 Particle motions in poroelastic wavefields 71
2.4.3 Fluid flow and attenuation of seismic waves 72
2.4.4 Slow wavefields in the low-frequency range 73
2.4.5 Elastic P- and S-waves in the low-frequency range 75
2.5 The quasi-static approximation of poroelasticity 76
2.6 Sources of fluid mass and forces in poroelastic media 80
2.6.1 Fluid injection at a point of a poroelastic continuum 83
2.7 Boundary loading of poroelastic media 86
2.7.1 Loading of a poroelastic half-space by a fluid reservoir 87
2.7.2 Fluid injection into a spherical cavity of a poroelastic
continuum 89
2.8 Stress and pressure coupling for radially symmetric fluid
sources 95
2.9 Several non-linear effects of poroelastic deformations 100
2.9.1 Deformation of the pore and fracture space 101
2.9.2 Stiff and compliant porosities 103
2.9.3 Stress dependence of elastic properties 106
2.9.4 Non-linear nature of the Biot–Willis coefficient α 110
2.9.5 Magnitude of the poroelastic-stress coupling 111
2.9.6 Effective-stress coefficients 113
2.9.7 Stress dependence of permeability 114
2.10 Appendix. Reciprocity-based relationship between
compliances of porous media 116
3 Seismicity and linear diffusion of pore pressure 118

3.1 Case study: KTB 119
3.2 Linear relaxation of pore pressure as a triggering mechanism 124
3.3 Triggering front of seismicity 127
3.4 Seismicity fronts and poroelastic coupling 131
3.5 Seismicity and hydraulic anisotropy 134
3.6 Seismicity in hydraulically heterogeneous media 137
3.6.1 Eikonal-equation approach 139
3.6.2 Validity domain of the eikonal-equation approach 141
3.6.3 Effective-medium approach 143
3.7 Back front of seismicity 147
Contents vii
3.8 Strength of pre-existing fractures 150

3.8.1 Statistics of rock criticality 150
3.8.2 Case study: Soultz-sous-Forêts 154
3.8.3 Case study: Fenton Hill 157
3.9 Spatial density of seismicity 158
4 Seismicity induced by non-linear fluid–rock interaction 164
4.1 Seismicity induced by hydraulic fracturing 165
4.1.1 Triggering front of hydraulic fracturing 165
4.1.2 Back front of hydraulic fracturing 168
4.1.3 Case study: Cotton Valley 169
4.1.4 Estimating permeability of virgin reservoir rocks 176
4.1.5 Estimating permeability of hydraulic fractures 177
4.1.6 Case study: Barnett Shale 177
4.2 Seismicity induced by non-linear pressure diffusion 180
4.2.1 Non-linear diffusion and triggering fronts 183
4.2.2 Triggering fronts and diffusivity estimates 189
4.3 The model of factorized anisotropy and non-linearity 196
5 Seismicity rate and magnitudes 201
5.1 The model of point-like induced events 202
5.1.1 Event number and event rate during a monotonic
injection 203
5.1.2 Seismicity rate after termination of an injection 206
5.1.3 Case studies of magnitude distributions 208
5.1.4 Seismogenic index 215
5.1.5 Occurrence probability of events with given magnitudes 220
5.2 Statistics of large magnitudes 224
5.2.1 Observations 224
5.2.2 Statistics of earthquakes with finite rupture surface 229
5.2.3 Rupture-surface probability and geometry of
stimulated volumes 232
5.2.4 Distributions of magnitudes and the
Gutenberg–Richter law 236
5.2.5 Lower and upper bounds for magnitude distributions 239
5.2.6 Case studies of magnitude distributions and stress
drop 243
5.2.7 Induced and triggered events 250
5.2.8 Maximum magnitude and sizes of stimulated volume 250
5.3 Appendix 1. Probability of discs within a sphere 253
5.4 Appendix 2. Probability of discs within an ellipsoid 254
viii Contents
5.5 Appendix 3. Probability of discs with centers inside a sphere 256

5.6 Appendix 4. Probability notations used in this chapter 258
References 259
Index 272
Color plate section between pages 82 and 83

Preface
Characterization of fluid-transport properties of rocks is one of the most important,

yet one of most challenging, goals of reservoir geophysics. However, active seis-
mic methods have low sensitivity to rock permeability and mobility of pore fluids.
On the other hand, it would be very attractive to have the possibility of exploring
hydraulic properties of rocks using seismic methods because of their large pene-
tration range and their high resolution. Microseismic monitoring of borehole fluid
injections is exactly the tool that can provide us with such a possibility. Borehole
fluid injections are often applied for stimulation and development of hydrocarbon
and geothermal reservoirs. Production of shale gas and heavy oil as well as CO2
sequestration are relatively recent technological areas that require broad applica-
tions of this technology. The fact that fluid injection causes seismicity has been well
established for several decades (see, for example, Pearson, 1981, and Zoback and
Harjes, 1997). Current ongoing research is aimed at quantifying and controlling
this process. Understanding and monitoring of fluid-induced seismicity is neces-
sary for hydraulic characterization of reservoirs and for assessments of reservoir
stimulations.
Fluid-induced seismicity covers a wide range of processes between the two
following limiting cases. In liquid-saturated hard rocks with low to moderate per-
meability (10−5 –10−2 darcy) and moderate bottom hole injection pressures (as a
rule, less than the minimum absolute value of the principal compressive tectonic
stress) the phenomenon of microseismicity triggering is often caused by the pro-
cess of linear relaxation of pore-pressure perturbations (Shapiro et al., 2005a,b).
Note that we speak here about the linearity in the sense of corresponding differ-
ential equations. In porodynamics this process corresponds to the Frenkel–Biot
slow wave propagation (see Biot, 1962, and a history review by Lopatnikov
and Cheng, 2005, as well as an English translation of Frenkel, 2005). In the
porodynamic low-frequency range (hours or days of fluid-injection duration) this
process reduces to a linear pore-pressure diffusion. Then, the linear pore-pressure
ix
x Preface
diffusion defines features of the rate of spatial growth, geometry and density of
clouds of microearthquake hypocenters (Shapiro et al., 2002, 2003, 2005a,b; Paro-
tidis et al., 2004). In some cases, spontaneously triggered natural seismicity, like
earthquake swarms, also shows similar diffusion-like signatures (Parotidis et al.,
2003, 2004, 2005; Hainzl et al., 2012; Shelly et al., 2013).
Another extreme case is a strong non-linear fluid–solid interaction related to the
hydraulic fracturing of sediments like a tight sandstone or a shale with extremely
low permeability (10−9 –10−5 darcy). In this case a fluid injection leads to a
strong enhancement of the permeability. Propagation of a hydraulic fracture is
accompanied by opening of a new fracture volume, fracturing fluid loss and its
infiltration into reservoir rocks, as well as diffusion of the injection pressure into
the pore space of surrounding formations and inside the hydraulic fracture (Econo-
mides and Nolte, 2003). Some of these processes can be seen from features of
spatio-temporal distributions of the induced microseismicity (Shapiro et al., 2006b;
Fischer et al., 2008; Dinske et al., 2010). The initial stage of fracture volume
opening as well as the back front of induced seismicity (propagating after ter-
mination of the fluid injection) can be observed. Evaluation of spatio-temporal
dynamics of induced microseismicity can help to estimate physical characteristics
of hydraulic fractures, e.g. penetration rate of the fracture, its permeability as well
as the permeability of the reservoir rock. Therefore, understanding and monitor-
ing of fluid-induced seismicity by hydraulic fracturing can be useful for describing
hydrocarbon and geothermal reservoirs and for estimating the results of hydraulic
fracturing.
Seismicity induced by borehole fluid injections is a central topic of this book. It
describes physical fundamentals of interpretation of fluid-induced seismicity. The
first two chapters of the book provide readers with an introduction to the theoreti-
cal background of concepts and approaches useful for understanding fluid-induced
seismicity. An application-interested reader can probably skip these two chapters
and just go directly to Section 1.4 and then Chapters 3–5, using Chapters 1 and 2
mainly as reference material.
In Chapter 1 the book starts with a brief introduction to the theory of elas-
ticity and seismic-wave propagation. This chapter also includes elements of
fracture mechanics and of the geomechanics of faulting. Then there is an intro-
ductory description of earthquake sources of the seismic wavefield. Finally, the
chapter contains a brief schematic description of methodical approaches of micro-
seismic monitoring. Many important processing-related methodical aspects of
microseismic monitoring remain outside of the scope of this book.
Chapter 2 provides a detailed introduction to the theory of poroelasticity. The
main physical phenomena responsible for fluid-induced seismicity and discussed in
this book in detail are fluid filtration and pore-pressure relaxation. They are closely
Preface xi
related to slow waves in porous fluid-saturated materials. The dynamics of slow

wavefields is the focus of this chapter. The chapter also includes a discussion of
some non-linear effects related to deformations of the pore space. They are relevant
for characterizing poroelastic coupling and for formulating models of the pressure-
dependent permeability. Such models will be used for the considerations of non-
linear pressure diffusion in subsequent chapters. The topic of thermo-poroelastic
interaction is not discussed in the book.
In Chapters 3–5 of this book we describe the main quantitative features of dif-
ferent types of fluid-induced microseismicity. Different properties of induced seis-
micity related to reservoir characterization and hydraulic fracturing are addressed,
along with the magnitude distribution of seismicity induced by borehole fluid injec-
tions. Evidently, this is an important question closely related to seismic hazard
of injection sites. Many corresponding aspects of the book are also applicable to
induced tectonic seismicity.
This book attempts to contribute to further elaboration of the seismicity-based
reservoir characterization approach (see also Shapiro, 2008).
Acknowledgments
This book contains results of research funded by different institutions in differ-

ent time periods. This includes the German Federal Ministry for the Environment,
Nature Conservation and Nuclear Safety (BMU), and the German Federal Min-
istry of Education and Research (BMBF) supported the section of Geophysics
of the Freie Universität Berlin for their projects MAGS and MEPRORISK. This
also includes the Deutsche ForschungsGemeinschaft (DFG) in whose Heisenberg
research program I started to work, in 1997, with interpretation of microseismic
monitoring of fluid injection at the German KTB. At that time I worked at the
Geophysical Institute of the Karlsruhe University and also spent a short period
at the GeoForschungsZentrum Potsdam. I continued this research at the Geolog-
ical School of Nancy, France, in close cooperation with Jean-Jacques Royer and
Pascal Audigane, where our work was significantly supported by the GOCAD
consortium project led by Professor Jean-Laurent Mallet. In 1999 I moved to the
Freie Universität Berlin. Many of the results reported here were then obtained in
our common work with Elmar Rothert, Jan Rindschwentner, Miltiadis Parotidis,
Robert Patzig, Inna Edelman, Nicolas Delepine and Volker Rath. Correspond-
ing research works were funded to a significant extent by the Wave Inversion
Technology (WIT) university consortium project led by Professor Peter Hubral
of Karlsruhe University.
Starting in 2005, the research reported here was to a large extent funded by
the PHysics and Application of Seismic Emission (PHASE) university consor-
tium project at the Freie Universität Berlin. Susanne Rentsch, Carsten Dinske,
Jörn Kummerow, Stefan Buske, Erik Saenger, Stefan Lüth, Cornelius Langen-
bruch, Oliver Krüger, Nicolas Hummel, Anton Reshetnikov, Antonia Oelke, Radim
Ciz, Maximilian Scholze, Changpeng Yu, Sibylle Mayr and Karsten Stürmer con-
tributed strongly to the work performed in this project. Corresponding results are
of especial importance for this book. I express my sincere thanks to all these
colleagues and friends, and I would like to thank the sponsors of the PHASE
xii
Acknowledgments xiii
consortium project and of two other consortium projects mentioned above as well
as the DFG, the BMU and the BMBF for their generous support of my work related
to this book.
I am also indebted to colleagues and institutions who helped me to access dif-
ferent microseismic data sets used here. These are Hans-Peter Harjes (Bochum
University),1 André Gérard and Roy Baria (SOCOMINE), Andrew Jupe (EGS
Energy), Michael Fehler and James Rutledge (LANL), Shawn Maxwell (Pinnacle
Technology), Kenneth Mahrer (USBR), Ted Urbancic, Adam Baig and Andreas
Wuesterfeld (ESG), Hideshi Kaieda (Central Research Institute of Electric Power
Industry, Tohoku), Takatoshi Ito (Institute of Fluid Science, Tohoku), Günter
Asch (GFZ-Potsdam), Martin Karrenbach (P-GSI), Ulrich Schanz and Markus
Häring (Geothermal Explorers), Sergey Stanchits and Georg Dresen (Deutsches
GeoForschungsZentrum, GFZ-Potsdam).
I am deeply grateful to colleagues who provided me with their comments on
the book manuscript. These are Boris Gurevich from Curtin University and CSIRO
(Perth) and Robert Zimmerman from Imperial College London (the first two chap-
ters). These are also my colleagues from the Freie Universität Berlin: Carsten
Dinske (Chapter 4), Jörn Kummerow (various sections), Cornelius Langenbruch
(Chapter 3), Oliver Krüger (Chapter 5). Of course, I have sole responsibility for
the complete book content.
I sincerely acknowledge the Society of Exploration Geophysicists (SEG) for
extending permission to use materials from a series of publications coauthored by
me in Geophysics and The Leading Edge.
1 Here, and in the following, the affiliations are given for the time periods during which the access to the data
was made possible.
1
Elasticity, seismic events and microseismic monitoring
By “seismic events” we understand earthquakes of any size. There exists a broad

scientific literature on earthquakes and on the processing of seismologic data. We
refer readers interested in a detailed description of these subjects to correspond-
ing books (see, for example, Lay and Wallace, 1995, and Shearer, 2009). We
start this book with an introductory review of the theory of linear elasticity and
of the mechanics of seismic events. The aim of this chapter is to describe classi-
cal fundamentals of the working frame necessary for our consideration of induced
seismicity. We conclude this chapter with a short introduction to methodical aspects
of the microseismic monitoring.
1.1 Linear elasticity and seismic waves

Deformations of a solid body are motions under which its shape and (or) its size
change. Formally, deformations can be described by a field of a displacement
vector u(r). This vector is a function of a location r of any point of the body in an
initial reference state (e.g., the so-called unstrained configuration; see, for exam-
ple, Segall, 2010). Initially we accept here the so-called Lagrangian formulation,
i.e. we observe motions of a given particle of the body.
However, the field of displacements describes not only deformations of the body
but also its possible rigid motions without changes of its shape and its size, such as
translations and/or rotations.
In contrast to rigid motions, under deformations, distances (some or any)
between particles of the body change. Therefore, to describe deformations, a
mathematical function of the displacement field is used that excludes rigid motions
of a solid and describes changes of distances between its particles only. This func-
tion is the strain tensor , which is a second-rank tensor with nine components i j .
Here the indices i and j can accept any of values 1, 2 and 3 denoting the coordinate
directions of a Cartesian coordinate system in which the vectors u and r have been
defined.
1
2 Elasticity, seismic events and microseismic monitoring
1.1.1 Strain
In the case of small deformations (i.e. where absolute values of all spatial deriva-
tives of any components of the vector u(r) are much smaller than 1) the strain
tensor has the form of a 3 × 3 symmetric matrix with the following components:

1 ∂u i ∂u j
i j = + . (1.1)
2 ∂x j ∂ xi
This form of the strain tensor describes deformations within a small vicinity of a
given location. This form remains the same also by consideration of small defor-
mations in the Eulerian formulation (see Segall, 2010), where instead of a motion
of a given particle of the body (i.e. the Lagrangian approach) rather a motion at
a given coordinate location (i.e. at a given point of the space) is considered. In
this book we accept the small-deformation approximation and do not distinguish
between the Lagrangian and Eulerian approaches.
Strains i j can be arbitrary (small) numbers. However, because of their definition
(1.1) they cannot be arbitrarily distributed in space. Spatial derivatives of strains
must be constrained by the following compatibility equations (see Segall, 2010):
∂ 2 i j ∂ 2 kl ∂ 2 ik ∂ 2 jl
+ = + . (1.2)
∂ x k ∂ xl ∂ xi ∂ x j ∂ x j ∂ xl ∂ xk ∂ xi
Deformations of a body results from applications of loads to it. Deformations
that will disappear completely if the loads are released are called elastic. Bodies
that can have elastic deformations are called elastic bodies.
1.1.2 Stress
Elastic bodies resist their elastic deformations by means of elastic forces. Elastic
forces in a solid body are analogous to a pressure in an ideal fluid. They occur due
to mutual interactions of elastically deformed parts of the body. These interactions
in turn take place on surfaces where the parts of the body are contacting each other
(see also Landau and Lifshitz, 1987).
Let us consider an elementary part of a body under deformation (see Figure 1.1).
Other parts of the body act by means of elastic forces onto this elementary part over
its surface S. Let us consider a differentially small element of this surface at its
arbitrary point r. Such a surface element can be approximated by a differentially
small part of a plane of area d S tangential to S at point r with a unit normal n
directed outside this part of the surface. Owing to elastic deformations an elastic
force dF(r, n) (also called a stress force) acts on the plane element with the normal
n. The following limit defines a traction vector:
x3 n
dS
dF(r,n)
S
r
x2
x1
Figure 1.1 A sketch for defining a traction.
dF(r, n)
τ (r, n) = lim . (1.3)
d S→0 dS
Note that the traction has the same physical units as a pressure in a fluid (e.g. Pa
in the SI system). Note also that the traction is a function of a location r and of an
orientation of the normal n.
Let us consider three plane elements parallel to coordinate planes at a given
location. We assume also that their normals point in the positive directions of
coordinate axes, which are perpendicular to the plane elements. Therefore, the cor-
responding three normals coincide with the unit basis vectors x̂1 , x̂2 and x̂3 of the
Cartesian coordinate system under consideration. Tractions acting on these plane
elements are τ (r, x̂1 ), τ (r, x̂2 ) and τ (r, x̂3 ), respectively. A 3 × 3 matrix composed
of nine coordinate components of these tractions defines the stress tensor, σ . Its
element σi j denotes the ith component of the traction acting on the surface with
the normal x̂j :
σi j = τi (x̂ j ). (1.4)
Let us consider a differentially small elastic body under an elastic strain and
assume for all deformation processes enough time to bring parts of this body into
an equilibrium state. From the equilibrium conditions for the rotational moments
(torques) of elastic forces it follows that the stress tensor is symmetric:
σi j = σ ji . (1.5)
Note that if the body torques are negligible (which is usually the case) this relation
is valid even in the case of the presence of rotational motions. This is because
of the fact that, in the limit of a small elementary volume, the inertial forces are
decreasing faster than the elastic force torques (see Auld, 1990, volume 1, section
2, for more details).
Similarly, a consideration of forces (elastic forces, body forces and inertial
forces) acting on a volume element in the limit of its vanishing volume shows
that elastic forces applied to the surface of such a volume must be in balance (see
Auld, 1990, volume 1, section 2, for more details). It then follows that a traction
τ (r, n) acting on an arbitrarily oriented plane surface element can be computed by
using the stress tensor:
τi (n) = σi j n j . (1.6)
Note that here and generally in this book (if not specially mentioned) we accept
the agreement on summation on repeated indices, e.g. ai bi = a1 b1 + a2 b2 + a3 b3 .
Definition (1.4) of the stress tensor corresponds to a common continuum
mechanics sign convention that tensile stresses are positive and compressive
stresses are negative (see, for example, a thin elementary volume and tractions
acting on its outer surface with normals pointing outside this volume; Figure 1.2).
1.1.3 Stress–strain relations

The strain-tensor and stress-tensor notations give a general form of an observa-
tional fact, known as Hooke’s law, that small elastic deformations are proportional
to elastic forces:
x3
dB
n(B–dB) n(B)
τ(B–dB) B τ(B)
x2
x1
Figure 1.2 A sketch illustrating positiveness of tensile stresses. Indeed,

equation (1.6) requires that the components σ22 in the both points, B and B − d B
must be positive. Note that the point B is shown as a dot on the right-hand side
of the disc. The point denoted as B − d B is not seen. It is on the left-hand side of
the disc; d B denotes the width of the disc.
i j = Si jkl σkl , (1.7)
where the fourth-rank tensor S, with components Si jkl , is the tensor of elastic com-
pliances. Note that their physical units are inverse to the unit of stress: 1/Pa. Owing
to the symmetry of the strain and stress tensors, the tensor of elastic compliances
has the following symmetries:
Si jkl = S jikl = Si jlk . (1.8)
Another fourth-rank tensor C, with components Ci jkl , called the tensor of elastic
stiffnesses, yields an alternative formulation of Hooke’s law:
σi j = Ci jkl kl . (1.9)
From this equation it is clear that the tensor of elastic stiffnesses also has the
symmetry:
Ci jkl = C jikl = Ci jlk . (1.10)
Both the tensor of elastic stiffnesses and the tensor of elastic compliances are
physical characteristics of a given elastic body.
Often both forms of Hooke’s law (1.7) and (1.9) are written symbolically as (see
Auld, 1990):
= S : σ , σ = C : . (1.11)
Here the double-dot (or double scalar) products denote summations over pairs of
repeating indices in (1.7) and (1.9), respectively.
A deformed elastic body possesses an elastic strain energy. At zero strain this
energy is equal to zero. With increasing strain by an increment dkl due to the stress
σkl , the volumetric density of this energy (energy per unit volume) must increase
by the increment d E = σkl dkl (see Landau and Lifshitz, 1987). The tensor of
elastic stiffnesses can then be used to define the density of the elastic strain energy
(by integration of the increment d E) as a positive quadratic function of non-zero
strains:
1 1 1
E = Ci jkl i j kl = σkl kl = Skli j σkl σi j , (1.12)
2 2 2
where in the two last expressions the two forms of Hooke’s law (1.7) and (1.9) have
been used. The product i j kl remains unchanged if the index pair i j is replaced by
kl and kl is replaced by i j, respectively. Thus, the tensor of elastic stiffnesses as
well as the the tensor of compliances must also have the following symmetry:
Ci jkl = Ckli j , Si jkl = Skli j . (1.13)

Symmetries (1.8), (1.10) and (1.13) of the stiffness and compliance tensors
reduce the number of their independent components. From 81 possible compo-
nents of a tensor (tensors’ indices can be equal to 1, 2 or 3) only 21 components
are mutually independent. These components are also called elastic moduli or
elastic constants (the latter notation neglects such effects as pressure dependence
and temperature dependence of these quantities). The requirement that the elastic
strain energy must be a positive-definite quadratic form of arbitrary strain/stress
components (called also the stability condition) provides additional restrictions on
allowed values of elastic moduli.
1.1.4 Elastic moduli

The tensors Ci jkl and Si jkl are inverse to each other so that (see Cheng, 1997):
1
Ci jkl Sklmn = (δim δ jn + δin δ jm ), (1.14)
2
where quantity δkl is the so-called Kronecker matrix, with components δkl = 1, for
k = l, and δi j = 0 in other cases.
The tensors of stiffnesses and compliances can be expressed in convenient matrix
forms by using their 21 independent components, respectively. For this, one uses
the so-called contracted notation (or the Voigt notations). Let us introduce capital
indices (e.g. I, J, etc.), which can take values 1, 2, 3, 4, 5 and 6. The following
correspondence between the capital indices and the pairs of the usual indices (i j)
is assigned: 1 → 11, 2 → 22, 3 → 33, 4 → 23, 5 → 13, and 6 → 12. In these
notations Hooke’s law has the following forms (Jaeger et al., 2007; Auld, 1990):
⎡ ⎤ ⎡ ⎤⎡ ⎤
11 s11 s12 s13 s14 s15 s16 σ11
⎢ ⎥ ⎢s ⎥⎢ ⎥
⎢ 22 ⎥ ⎢ 12 s22 s23 s24 s25 s26 ⎥ ⎢σ22 ⎥
⎢ ⎥ ⎢ ⎥⎢ ⎥
⎢ 33 ⎥ ⎢s13 s23 s33 s34 s35 s36 ⎥ ⎢σ33 ⎥
⎢ ⎥=⎢ ⎥⎢ ⎥, (1.15)
⎢223 ⎥ ⎢s14 s24 s34 s44 s45 s46 ⎥ ⎢σ23 ⎥
⎢ ⎥ ⎢ ⎥⎢ ⎥
⎣213 ⎦ ⎣s15 s25 s35 s45 s55 s56 ⎦ ⎣σ13 ⎦
212 s16 s26 s36 s46 s56 s66 σ12
⎡ ⎤ ⎡ ⎤⎡ ⎤
σ11 c11 c12 c13 c14 c15 c16 11
⎢σ ⎥ ⎢c ⎥⎢ ⎥
⎢ 22 ⎥ ⎢ 12 c22 c23 c24 c25 c26 ⎥ ⎢ 22 ⎥
⎢ ⎥ ⎢ ⎥⎢ ⎥
⎢σ33 ⎥ ⎢c13 c23 c33 c34 c35 c36 ⎥ ⎢ 33 ⎥
⎢ ⎥=⎢ ⎥⎢ ⎥. (1.16)
⎢σ23 ⎥ ⎢c14 c24 c34 c44 c45 c46 ⎥ ⎢223 ⎥
⎢ ⎥ ⎢ ⎥⎢ ⎥
⎣σ13 ⎦ ⎣c15 c25 c35 c45 c55 c56 ⎦ ⎣213 ⎦
σ12 c16 c26 c36 c46 c56 c66 212
In these two equations the contracted notation is used in the two symmetric 6 × 6
matrices of components s I K and c I K , where I corresponds to a pair of normal
indices, e.g. i j, and K corresponds to another their pair, e.g. kl. It is clear that
these matrices are inverse to each other, i.e. their matrix product gives a 6 × 6
unit matrix. Their components are also called elastic compliances and elastic stiff-
nesses, respectively. The relations between the contracted-notation stiffness matrix
components and corresponding components of the fourth-rank tensor of elastic
stiffnesses is simple: c I K = Ci jkl . This correspondence for the compliances is a
bit more complicated: s I K = Si jkl if I, K = 1, 2, 3, s I K = 2Si jkl if I = 1, 2, 3 and
K = 4, 5, 6, and s I K = 4Si jkl if I, K = 4, 5, 6.
The higher the physical symmetry of the elastic medium, the smaller is the num-
ber of non-vanishing independent elastic constants. For mineral crystals, different
symmetries are of relevance (see Auld, 1990, for a comprehensive description). In
the most general case of triclinic crystals the elastic properties are characterized by
21 independent compliances (or, equivalently, 21 independent stiffnesses). This sit-
uation corresponds to equations (1.15) and (1.16), respectively. If the medium has a
single symmetry plane (the monoclinic symmetry) then the number of independent
constants will be reduced to 13 (for example, if we assume the x y coordinate plane
as the plane of symmetry, this will result in the invariant coordinate transformation
z → −z and thus, all elastic constants with odd numbers of index 3 must be equal
to zero). This situation corresponds, for example, to a layered medium with a single
system of plane cracks oblique to the lamination plane.
One of most relevant symmetries for rocks is the orthorhombic one. It can be
applied to describe different geological situations, like rocks with three mutu-
ally perpendicular systems of cracks or horizontally layered rocks permeated by
a single system of aligned vertical fractures. An orthorhombic medium has three
mutually perpendicular symmetry planes. This means that in such a medium under
corresponding coordinate transformations (reflections across symmetry planes) the
tensors of elastic constants must remain unchanged. In a coordinate system with
axes normal to the symmetry planes it follows that all components Ci jkl and Si jkl
with odd numbers of any index must be equal to zero. This leads to the following
forms of the compliance and stiffness matrices, respectively:
⎡ ⎤ ⎡ ⎤
s11 s12 s13 0 0 0 c11 c12 c13 0 0 0
⎢s 0⎥ ⎢ 0⎥
⎢ 12 s22 s23 0 0 ⎥ ⎢c12 c22 c23 0 0 ⎥
⎢ ⎥ ⎢ ⎥
⎢ 13 23 33
s s s 0 0 0 ⎥ ⎢ 13 23 33
c c c 0 0 0 ⎥
⎢ ⎥; ⎢ ⎥ . (1.17)
⎢0 0 0 s44 0 0⎥ ⎢0 0 0 c44 0 0⎥
⎢ ⎥ ⎢ ⎥
⎣0 0 0 0 s55 0 ⎦ ⎣ 0 0 0 0 c55 0 ⎦
0 0 0 0 0 s66 0 0 0 0 0 c66
We see that nine independent constants are enough to completely describe the
elastic properties of an orthorhombic medium. The compliances can be obtained
from stiffnesses by the matrix inversion and vice versa. In the case of an arbitrary
coordinate orientation, three additional constants (corresponding to three rotational

angles) are required.
A useful and geologically relevant subset of orthorhombic symmetry is
transverse isotropy. Layered sedimentary rocks can frequently be described by
this symmetry. The plane of lamination is then the symmetry plane. If one of
the coordinate planes coincides with the symmetry plane, then a coordinate axis
normal to the symmetry plane will be an axis of an arbitrary-angle rotational
symmetry. This symmetry results in four additional relations between the elastic
constants, reducing the number of independent ones to five. If the symmetry axis
coincides with the direction of the axis x3 , then in equations (1.17) additional rela-
tions will be (Auld, 1990): s22 = s11 , s23 = s13 , s55 = s44 and s66 = 2(s11 − s12 ).
Correspondingly, c22 = c11 , c23 = c13 , c55 = c44 and c66 = (c11 − c12 )/2.
Finally, in the case of an elastic isotropic medium (all coordinate axes are
arbitrary-angle rotational symmetry axes and any plane is a plane of symmetry),
two constants remain independent only: s22 = s33 = s11 , s23 = s13 = s12 ,
s66 = s55 = s44 and s44 = 2(s11 − s12 ). Correspondingly, c22 = c33 = c11 ,
c23 = c13 = c12 , c66 = c55 = c44 and c44 = (c11 − c12 )/2. The independent elastic
stiffnesses are usually denoted as the elastic moduli λ and μ, so that c44 = μ and
c12 = λ. Inverting the matrix ci j we obtain compliances of an isotropic medium:
λ+μ λ 1
s11 = , s12 = − , s44 = . (1.18)
μ(3λ + 2μ) 2μ(3λ + 2μ) μ
Let us consider a volumetric strain (dilatation) of an elementary volume V of an

arbitrary anisotropic elastic medium:
dV
≡ . (1.19)
V
We can choose such an elementary volume to be a cuboid with side lengths l x , l y
and l z . Thus we see that
d(l x l y l z ) dl x dl y dl z
= = + + = 11 + 22 + 33 . (1.20)
lx l y lz lx ly lz
Let us further assume that this dilatation is a result of a hydrostatic stress,

σkl = − pδkl , applied to the medium, where p is the pressure loading the medium.
A general relation between the dilatation and the pressure can be obtained by tak-
ing a double-dot product (the scalar product) of Hooke’s law (1.7) with the δi j (i.e.
multiplying the both sides with δi j and summing up over repeating indices):
= −Siikk p. (1.21)
The proportionality coefficient here is a bulk compressibility C mt of the elastic

material:
C mt ≡ Siikk = S1111 + S2222 + S3333 + 2(S1122 + S1133 + S2233 )
= s11 + s22 + s33 + 2(s12 + s13 + s23 ). (1.22)
It follows from (1.19)–(1.21) that the bulk compressibility of a sample has the
following relation to its bulk density ρ:
dV d(1/ρ) 1 dρ
C mt = − =− = . (1.23)
V dp (1/ρ)dp ρ dp
In the case of an isotropic elastic material we obtain (see equations (1.22) and
(1.18)) C mt = 3s11 + 6s12 = 1/(λ + 2μ/3). Therefore,
K = λ + 2μ/3 (1.24)
is a bulk modulus describing the stiffness of the material to volumetric
deformations.
The following representation of the stiffness tensor of an isotropic medium is
useful (Aki and Richards, 2002):
Ci jkl = λδi j δkl + μ(δik δ jl + δil δ jk ). (1.25)
In the same terms, Hooke’s law for isotropic elastic media can be written in the
following form:
σi j = λδi j + 2μi j . (1.26)
From this equation it follows that μ is a shear modulus of the material, describing
its stiffness to shear deformations (under which i = j). It follows also that under
uniaxial stress conditions (for example σ33 = 0 and σ11 = σ22 = 0) the ratio ν of
the transverse strain to the longitudinal strain, −11 /33 , is equal to
λ
ν= . (1.27)
2(λ + μ)
This quantity is called Poisson’s ratio. For an isotropic elastic solid the stability
condition requires that both bulk and shear moduli must be positive. For Poisson’s
ratio this yields the restriction −1 ≤ ν ≤ 0.5. For realistic rocks this coefficient is
positive. Its upper limit of 0.5 corresponds to fluids. Frequently, its values for stiff
tight isotropic rocks are close to 0.25 (corresponding to λ ≈ μ).
All elastic moduli introduced above will usually be assumed to be isothermal
ones, if static deformations or processes being very slow in respect to the thermal
diffusion are considered. In this book we consider processes that are faster than the
temperature equilibration (e.g. wave propagation and pore-pressure equilibration).
We will assume that these processes are approximately adiabatic. Thus, we assume
that the elastic moduli introduced above are adiabatic. Note that the adiabatic and
isothermal moduli of hard materials (e.g. rocks) differ by a small amount (see also
Landau and Lifshitz, 1987).
In this book we will frequently assume that the elastic properties of the medium
are isotropic. This simplifying assumption is often too rough for problems of
seismic event location and imaging (which are not the main subject of our
consideration). For such problems velocity models should take into account seis-
mic anisotropy at least in the weak anisotropy approximation (Thomsen, 1986;
Tsvankin, 2005; Grechka, 2009). For describing dominant effects responsible for
the triggering of induced microseismicity the assumption of elastic isotropy seems
to be adequate at least as the first approximation. For such effects hydraulic
anisotropy of rocks is much more important. Elastic anisotropy in rocks is usually
below 10% and seldom exceeds 30%, in respect to the velocity contrast between
the slowest and fastest wave propagation directions. In shale the elastic anisotropy
can be even higher. However, usually it is much smaller than a possible anisotropy
of the hydraulic permeability, which can reach several orders of magnitude.
1.1.5 Dynamic equations and elastic waves

By an elastic deformation, a transfer of an elastic solid from one equilibrium state
to another equilibrium state occurs by means of propagation of elastic waves.
Elastic waves in rocks in the frequency range between 10−3 and 104 Hz are usu-
ally referred to as seismic waves. Resulting elastic forces acting on an elementary
volume of the elastic medium define its acceleration vector. Owing to Hooke’s
law and the definition of the strain tensor, the second Newtonian law (i.e. the
momentum conservation) takes the form of the following dynamic equation (Lamé
equation):
∂ ∂u k ∂ 2ui
Ci jkl =ρ 2 . (1.28)
∂x j ∂ xl ∂t
This equation describes the propagation of elastic waves in the most general case
of a heterogeneous anisotropic elastic medium. Note that this is a system of three
equations for three unknown components of the displacement vector. A plane-
wave analysis (see also our later discussion of poroelastic waves) is instructive
for investigating modes of propagation of elastic perturbations.
Let us consider the case of a homogeneous arbitrary anisotropic elastic medium.
Then equation (1.28) simplifies to:
∂ 2uk ∂ 2ui
Ci jkl =ρ 2 . (1.29)
∂ x j xl ∂t
In linear systems any wavefield can be decomposed into a superposition of

independently propagating time-harmonic plane waves. We will designate their
angular frequencies by ω. Thus, we can look for a solution of equation (1.29) in
the following form:
u = u0 ei(ωt−κr) , (1.30)
Here κ is a wave vector of an arbitrary direction and an unknown length and u0 is

an unknown polarization vector of a plane wave. Substituting this expression into
equation (1.29) we obtain
Ci jkl u 0k κ j κl = ρω2 u 0i . (1.31)
Using the Kronecker matrix we can rewrite this equation in a more instructive way:
(Ci jkl κ j κl − ρω2 δik )u 0k = 0. (1.32)
This matrix equation is called the Christoffel equation. This is a homogeneous

system of three algebraic linear equations for the three components of the polariza-
tion vector u0 . Therefore, this equation system has non-zero solutions only if the
following determinant is equal to zero:
det(Ci jkl κ j κl − ρω2 δik ) = 0. (1.33)
Equation (1.33) defines a relationship between the wave vector and the frequency
for any possible plane wave in an elastic medium with the density ρ and the stiff-
ness tensor Ci jkl . Thus, it is called a dispersion equation. It is clear that (1.33) is an
equation for the following quantity:
|ω|
c= . (1.34)
|κ|
This quantity is the phase velocity of the corresponding plane wave. The condi-
tion (1.33) is a cubic equation in respect to c2 . Therefore, this equation describes
dispersion relations of three different types of elastic waves. In weakly anisotropic
media these waves correspond to an independently propagating quasi P- (nearly
longitudinal) and two different quasi S- (nearly shear) waves. Substituting any of
the three possible solutions for c2 into equation (1.32) one obtains solutions for the
corresponding polarization vectors u0 .
Equation (1.30) shows that the phase velocity c controls how quickly the argu-
ment (i.e. the phase) of a given time-harmonic plane wave changes in space and
time. In contrast, the equation
∂ω
gi = (1.35)
∂κi
provides components gi of the group velocity vector of the wave type corre-
sponding to the solution of equation (1.33) for c2 . The group velocity describes
propagation of a spatial envelope of a group of plane waves with differentially
close wave vectors. It is indeed a propagation velocity of wavefronts radiated by
finite sources in their far fields in anisotropic elastic media. In elastic media the
group velocity describes wave signal propagation along rays.
In the case of a homogeneous isotropic medium the phase and the group veloc-
ities coincide. In this case equation (1.33) provides two different solutions for
c2 only. They correspond to seismic P- and S-waves. Also a decomposition into
rotational and dilatational parts of the displacement field (an example of such a
decomposition will be shown later in our discussion of poroelastic wavefields)
shows that equation (1.28) describes these two independently propagating elastic
waves: the longitudinal P-wave and the shear S-wave with propagation velocities
c p and cs , respectively:
λ + 2μ μ
cp = , cs = . (1.36)
ρ ρ
Note that similarly to K and μ, the so-called P-wave modulus, c11 = λ + 2μ =
K + 43 μ, is always positive.
1.1.6 Point sources of elastic waves

Above we considered plane waves as elementary components of elastic wavefields.
Solutions of the linear equation system (1.28) can also be obtained as a super-
position of waves radiated by point sources. A rather general formulation of the
superposition principle for a linear system (including the linear equation system
of poroelasticity considered in the next chapter) is as follows. Let us assume that
we know a (possibly tensorial) solution G(r, r0 , t − t0 ) for a given field quantity
satisfying such a system with a source term equal to the Dirac function δ(r0 )δ(t0 )
(i.e. a singular source concentrated in point r0 and time moment t0 ). Such a solu-
tion is called a Green’s function. Let us further assume that, in reality, the source
terms are distributed in space and time, as a known source-density function f s (r, t)
describes. Then the corresponding field quantity produced by this source will be
equal to the following superposition:
∞
G(r, r0 , t − t0 ) f s (r0 , t0 )d 3 r0 dt0 . (1.37)
−∞
Sometimes it is convenient to work with source terms having Heaviside-type

temporal distributions h(t) (where h(t) = 0 if t < 0 and h(t) = 1 if t ≥ 0).
Note that such Green’s functions require their differentiation in respect to time (or
alternatively, temporal differentiation of the function f s (r, t)) before applying the
superposition (1.37).
Let us consider point-like distributions of instantaneous body-force densities in
an isotropic homogeneous elastic medium. To complete the force balance a body-
force source must be added to the left-hand part of the elastodynamic equation
(1.28). Such a singular source F acting in direction x̂j can be expressed in the
following form:
F(r, t) ∝ x̂j δ(r)δ(t). (1.38)
In the far field of this source (at the distance larger than several wavelengths) the
wavefield is given by the following approximate Green’s function (see chapter 4 of
Aki and Richards, 2002):
δ(t − r/c p ) δ(t − r/cs )
G i j ≡ u i (r, t) ∝ γi γ j 2
+ (δi j − γi γ j ) , (1.39)
4πρc p r 4πρcs2r
where r = |r|, and the quantities γi = ri /r are the direction cosines. The first term
on the right-hand part describes a spherical P-wave. The second one describes a
spherical S-wave.
1.1.7 Static equilibrium

Let us assume that elastic waves have had enough time to establish an equilibrium
state in a given body. Then displacement vectors and strains will become indepen-
dent of time. Therefore, an equation of elastic equilibrium can be obtained from
equation (1.28) by removing the temporal derivatives of the displacements (i.e. by
taking the static limit):
∂ ∂u k
Ci jkl = 0. (1.40)
∂x j ∂ xl
Note that in terms of the stress tensors
∂
σi j = 0. (1.41)
∂x j
Using equation (1.26) for homogeneous isotropic media we obtain:
∂ ∂i j
λ + 2μ = 0. (1.42)
∂ xi ∂x j
The second term on the left-hand side of this equation can be written explicitly as
follows:

∂i j ∂(∂u i /∂ x j + ∂u j /∂ xi ) ∂ 2ui ∂ 2u j
2μ =μ =μ + . (1.43)
∂x j ∂x j ∂ x 2j ∂ xi ∂ x j
The third part of this equation is equal to the ith component of the vector μ(∇ 2 u +
∇(∇u)), where we have introduced a differential vector operator
∇ ≡ (∂/∂ x1 ; ∂/∂ x2 ; ∂/∂ x3 ). (1.44)
Note that with this notation:
= ∇u. (1.45)
Here and later we use also the Laplace operator
∂ ∂ ∂2 ∂2 ∂2
∇2 ≡ ≡ 2 + 2 + 2. (1.46)
∂ xi ∂ xi ∂ x1 ∂ x2 ∂ x3
Therefore, equation (1.42) has the following vectorial form:
(λ + μ)∇ + μ∇ 2 u = 0, (1.47)
Further we apply a known relation of the vector calculus:
∇(∇u) = ∇ 2 u + ∇ × (∇ × u). (1.48)
Using this we obtain
(λ + 2μ)∇ − μ∇ × (∇ × u) = 0. (1.49)
Applying here once more the divergence operator (defined by equation 1.44) we
obtain
∇ 2 = 0. (1.50)
Thus, we see that, in homogeneous isotropic media under equilibrium without body
forces, the dilatation is a harmonic function (i.e. it is a solution of the Laplace
equation). Also applying the Laplace operator directly to (1.47) yields:
∇ 2 ∇ 2 u = 0. (1.51)
This is the so-called biharmonic equation for the displacement vector under static
equilibrium conditions for an elastic medium (see Landau and Lifshitz, 1987).
1.2 Geomechanics of seismic events

Earthquakes are usually assumed to occur due to spontaneous quick shear motions
(shear failure) along tectonic faults and cracks. A rather simple and practical for-
mulation of conditions of such a brittle failure of rocks is based on Anderson’s
theory of faulting published in 1905. Anderson’s theory was in turn based on ear-
lier studies of Amontons in 1699 and Coulomb in 1773 on rock mechanics (Lay and
Wallace, 1995). This formulation follows from a consideration of tectonic stresses.
1.2.1 Faults and principal stresses

Let us consider a tensor of tectonic stresses in its principal coordinate system. In
such a coordinate system the traction vectors acting on the coordinate planes are
directed along the normals to these planes. In other words, these traction vectors
are parallel to the coordinate axes normal to the corresponding coordinate planes.
These traction vectors are usually called principal tectonic stresses.
Frequently in tectonics and geomechanics the following sign notation is used for
stresses: compressive stresses are positive, and a pore pressure is positive. How-
ever, in continuum mechanics it is usually assumed that compressive stresses are
negative, and the pore pressure is positive. As was noted in the previous sections,
in this book we accept the standard continuum mechanic sign-notation system, i.e.
compressive stresses (a usual tectonic load is compressive) must be negative.
We will denote the projections of the principal stresses onto the corresponding
(parallel to them) coordinate axes (note that each of the principal stresses has only
one such non-vanishing projection) by −σ1 , −σ2 and −σ3 , respectively. In this
notation, in the case of compressive stresses, the algebraic quantities σ1 , σ2 and σ3
are positive. We will order the principal stresses so that the quantities σ1 , σ2 and σ3
will be in the following relations:
σ1 ≥ σ2 ≥ σ3 . (1.52)
In the case of compressive stresses the quantities σ1 , σ2 and σ3 are equal to the
absolute values of the maximum, the intermediate and the minimum compressive
tectonic stresses, respectively.
Correspondingly, in the principal coordinate system, from the six independent
components of the stress tensor σi j only three (diagonal) ones are non-vanishing:
σx x , σ yy and σzz . It is usually observed that one of the principal tectonic stresses
is vertical. Then, x, y and z denote the two horizontal and one vertical coordi-
nate axes, respectively. For example, in the normal faulting regime the maximum
compressive stress σ1 is vertical, and thus:
σzz = −σ1 , σ yy = −σ2 , σx x = −σ3 . (1.53)
In the strike-slip faulting regime the maximum compressive stress σ 1 and the
minimum compressive stress σ 3 are both horizontal:
σzz = −σ2 , σ yy = −σ3 , σx x = −σ1 . (1.54)
Finally, in the thrust faulting regime the minimum compressive stress σ 3 is vertical:
σzz = −σ3 , σ yy = −σ2 , σx x = −σ1 . (1.55)

x3
n
θ
σ3 θ
θ
σ1 x1
Figure 1.3 A fault and the principal stresses. The fault is normal to the plane of
the maximum and minimum compressive stresses σ1 and σ3 . It is shown as the
line moving out at about 35◦ northeast from the origin. The compressive stresses
σ1 and σ3 are shown as segments of the coordinate axes x1 and x3 , respectively.
The three right triangles are mutually similar and have the same angle θ .
In these three stress regimes potential fault surfaces (planes) are parallel to the
direction of the intermediate principal compressive tectonic stress σ2 . Such a poten-
tial fault plane is normal to the plane of the maximum and minimum compressive
stresses (σ1 , σ3 ). Let us consider a potential fault plane making an angle θ with
the coordinate axis of the maximum compression σ1 (see Figure 1.3). A unit nor-
mal vector to this plane has the following components (note that we work in the
principal-stress coordinate system):
n 1 = − sin θ, n 2 = 0, n 3 = cos θ. (1.56)
Then a traction with the following components acts on this plane (we use equation
(1.6) and our notations for the projection of principal stresses):
τ1 = σ1 sin θ, τ2 = 0, τ3 = −σ3 cos θ. (1.57)
The component of this traction acting normal to the potential fault plane (we denote
it as −σn , see below) is given by a scalar product between the plane’s normal (1.56)
and the traction (1.57):
σ1 + σ3 σ1 − σ3
− σn = −σ1 (sin θ)2 − σ3 (cos θ)2 = − + cos(2θ). (1.58)
2 2
This quantity, −σn , is a normal stress acting onto the fault plane. Note that if
principal stresses are compressive then the normal stress is compressive too. Thus,
it is negative, and σn > 0. This explains our choice of the sign in front of σn in
equation (1.58).
A unit vector e located in plane (x; z) and directed along the fault plane has the
following components:
e1 = cos θ, e2 = 0, e3 = sin θ. (1.59)
Thus, the traction component τt tangential to the fault plane is

σ1 − σ3
τt = sin(2θ). (1.60)
2
This is a shear stress acting on the fault plane.
Note that from equations (1.58) and (1.60) it follows that

σ1 + σ 3 2 (σ1 − σ3 )2
σn − + τt2 = . (1.61)
2 4
This is the equation of a circle in the coordinate system (σn , τt ) with the center at
the point (σm ; 0), where
1
σm ≡ (σ1 + σ3 ) (1.62)
2
is the mean compressive stress. Equation (1.61) defines the so-called Mohr circle
(see Figure 1.4). The diameter of Mohr’s circle is equal to the differential stress:
σd ≡ σ1 − σ3 . (1.63)
The angle between a radius of the Mohr’s circle and the negative direction of the
σn -axis is equal to 2θ. The maximum possible shear stress on a fault plane is equal
to σd /2. It corresponds to the angle θ = 45◦ .
On the other hand, a friction-like force (per unit surface) acting along a potential
fault plane and keeping such a fault stable (i.e. non-sliding) is given by Amontons’
law:
τ f = μ f σn , (1.64)
where μ f is the friction coefficient. Note that, in the case of brittle failure of an
intact rock, this coefficient is addressed as a coefficient of internal friction. In
τf τt
τf (σn)
σd/2 τt (σn)
σ3 σm σ1 σn
Figure 1.4 The shear stress and the friction force per unit surface of a fault plane
as functions of the normal stress. Note that the center of the circle has coordinates
(σm ; 0). Its top point has coordinates (σm ; σd /2).
this case it is usually higher than the friction coefficient on a pre-existing fault
surface. Observation indicate that the induced seismicity occurs mainly on pre-
existing defects of rocks (see, for example, Zoback, 2010, and further chapters of
this book). Such defects (i.e. faults and cracks) will be destabilized if the absolute
value of the shear stress |τt | exceeds the friction τ f .
Frequently one expects that even in the case of a vanishing normal stress such
a destabilization still requires a finite amount of shear stress Cc . Thus, the failure
criterion (also called the Mohr–Coulomb failure criterion) for an arbitrary angle θ
can be formulated as follows:
|τt | = Cc + τ f . (1.65)
The quantity Cc is called the cohesive strength (or just the cohesion). In the case
of a failure of a pre-existing fault surface the cohesion is frequently neglected (see
chapter 4 of Zoback, 2010). Finally, in the presence of fluids in pores of rocks the
normal stress will be effectively reduced due to the pore pressure Pp and equation
(1.65) takes the following form (see also our discussion of the Terzaghi’s effective
stress in the next chapter after equation (2.280)):
|τt | = Cc + μ f (σn − Pp ). (1.66)
The pore pressure equally reduces all three principal stresses to corresponding
effective principal stresses. Thus, it is convenient to consider Mohr’s circle in the
new coordinate system (σn − Pp , τt ). Note that this coordinate transformation does
not change differential and shear stresses.
Let us firstly consider the half-space (σn − Pp , τt ) with positive shear stresses.
Equation (1.66), expressed in geometric terms, is shown in Figure 1.5. There are
three possible relative locations of the straight line
τ f c = Cc + μ f (σn − Pp ) (1.67)
expressing the sum of the cohesion stress with the friction force and of the semi-
circle

|τt | = (σ1 − σn )(σn − σ3 ) (1.68)
(see equation (1.61)) expressing the shear stress. They can have no intersections.
In this case the friction force is too large and no faults will be destabilized. Another
possibility is that the straight line and the semi-circle are intersect at two points. In
this case the shear stress is more than sufficient to destabilize faults having orien-
tations corresponding to the circle arc above the straight line. Finally, a situation
is possible where the straight line just touches the circle in a single point. This
means that the shear stress is exactly sufficient to destabilize faults of one special
(optimal) orientation. The corresponding angle θopt can be obtained from a simple
τf τt
Cc
σ3–Pp σ1–Pp σn–Pp
Figure 1.5 A friction–stress diagram taking into account a cohesion and a pore
pressure acting on faults. The solid lines show force and angle relations at opti-
mally oriented faults. The dashed lines show friction forces that are too large or
too small in respect to the one acting on the optimally oriented fault.
τf τt
A1 A2
A
σn–Pp
O S1 S2
Figure 1.6 A friction–stress diagram for compressive stresses. The solid lines
show force and angle relations for faults oriented at the lock-up angle. A doubled
lock-up angle is given by the angle OS1A. The dashed lines show deviating situ-
ations. They indicate that corresponding angles between the largest compressive
stress direction and faults (doubled these angles are given by the angles OS1A1
and OS2A2) are smaller than the lock-up angle.
geometric consideration. Recalling that in Figure 1.5 the friction coefficient μ f

gives a tangent of the friction straight line (i.e. μ f = tan φ f ) we obtain:
1 1
θopt = arctan . (1.69)
2 μf
Note that an equivalent consideration of the half-space (σn − Pp , τt ) with negative

shear stresses will provide the same angle but with the negative sign.
Usually the friction coefficient on a pre-existing fault surface is between 0.6 and
1. Thus, θopt = 30◦ − 22.5◦ . Figure 1.6 shows that for any compressive stress state
the faults with θ > 2θopt cannot slip. The quantity 2θopt is called the lock-up angle.
It is reported that faults with θ exceeding the lock-up angle are extremely rare (see
Kanamori and Brodsky, 2004).
At the optimal orientations of potential fault planes the normal and tangential
stresses can be explicitly expressed in terms of the friction coefficient. Indeed,
one can substitute the expressions cot(2θopt )= μ f = tan φ f , sin(2θopt ) =
1/ 1 + μ2f = cos φ f and cos(2θopt ) = μ f / 1 + μ2f = sin φ f into equations
(1.58) and (1.60):
σd
−σn = −σm + sin φ f . (1.70)
2
σd
τt = cos φ f . (1.71)
2
The Mohr–Coulomb failure criterion (1.66) at the optimal angle takes then the
following form:
1
σd = Cc cos φ f + sin φ f (σm − Pp ). (1.72)
2
The difference between the left- and right-hand sides of this equation represents a
failure criterion stress, FC S:
1
FC S = σd − sin φ f (σm − Pp ) − Cc cos φ f . (1.73)
2
If FC S is negative in the vicinity of a given location, then the rock will be
stable at this location. If changes of a failure criterion stress FC S are positive,
the rock will become less stable there. If the failure criterion stress becomes pos-
itive, the corresponding fault plane becomes unstable and brittle failure occurs:
the fragments of rocks contacting along this fault plane slip relative to each
other.
1.2.2 Friction coefficient

The Mohr–Coulomb criterion of rock failure is an approximation and simplifica-
tion of reality. For example, a time-independent coefficient of friction is a strong
assumption. The coefficient of internal friction introduced in the previous section
does not describe the friction force along an already (and possibly long geological
time) existing interface between two rock units. Theoretically the internal friction
coefficient addresses friction along a potential internal plane in intact rocks before
the failure occurs on this plane (Jaeger et al., 2007). On the other hand, realistically,
intact rocks do not exist, and failures probably occur along pre-existing zones of
weakness. Moreover, the empirically obtained mathematical forms of the failure
criterion of a plane of weakness and of the Mohr–Coulomb criterion coincide (see
Jaeger et al., 2007, equations 3.20 and 4.5). Thus, friction effects in both situations
must be at least similar.
A great deal of active research in modern seismology has been dedicated to

investigating the nature of the friction along pre-existing interfaces in rocks. A
comprehensive review on the subject can be found in the last two chapters of the
book of Segall (2010). The most important results are usually summarized by a
system of empirically derived equations (Dieterich, 1978; Ruina, 1983) describing
the dynamic friction coefficient μ f d along surfaces sliding with the slip velocity v
(Segall, 2010):
v θf
μ f d = μ f 0 + a f ln + b f ln , (1.74)
v0 θf0
where μ f 0 is a friction coefficient (frequently assumed to be close to 0.6) at a ref-
erence slip velocity v0 and at the quantity θ f (a so-called state variable) equal to its
reference value θ f 0 . The quantities a f and b f are empiric dimensionless constants
defining the behavior of the dynamic friction.
The state variable expresses the temporal behavior of the friction due to damage
accumulation at the friction interface. Alternatively, one of the following two dif-
ferential equations (they have also been empirically derived) describing the state
variable as a function of time can be used. The first one is called the aging law:
dθ f vθ f
=1− , (1.75)
dt dc
where dc stands for the critical slip required for the friction coefficient to obtain its
value by steady-state sliding after an abrupt change of the slip velocity. This model
proposes that the state variable is just linearly mapped to the time in the case of
v = 0 (i.e. a non-slipping contact).
Another model frequently used for describing the state variable is called the
slip law:

dθ f vθ f vθ f
=− ln . (1.76)
dt dc dc
Both of these models predict that, in steady-state sliding, dθ f /dt = 0 and
θ f = dc /v. Assuming that θ f 0 = dc /v0 they yield:
v
μ f d = μ f 0 + (a f − b f ) ln . (1.77)
v0
Thus, if the friction-law parameters a f and b f are such that a f > b f , then increas-
ing velocity will lead to increasing friction. The fault tends to become more stable.
This situation is called velocity strengthening. In the case of a f < b f an increas-
ing sliding velocity will decrease the friction and contribute to fault destabilization.
Such a situation is called velocity weakening.
Usually variations of the friction coefficient related to the state and rate processes
are considered to be of the order of 10%. They can be well hidden by the elastic
heterogeneity and stress variations in real rocks and contribute to the randomness
of the FC S distribution in space (and also in time).
1.2.3 Growth of finite cracks: a sufficient condition

Another simplification of the Mohr–Coulomb failure criterion is an assumption of a
homogeneous stress state in the whole continuum. Thus, a destabilization condition
of a part of a fault plane will be equally valid for the whole infinite corresponding
fault plane. However, in reality, sizes of rock failures are finite. A sufficient con-
dition of the unstable crack growth would be that a critical tensile stress must be
reached in the vicinity of the fracture tip.
Based on this assumption and considering a finite elliptical two-dimensional
(2D) initially tensile crack in the plane of the maximum and minimum stresses,
Griffith (1924) derived the following criterion (see the derivation given in section
10.9 of Jaeger et al., 2007):
(σ1 − σ3 )2 − 8T0 (σ1 + σ3 ) = 0 if σ1 > −3σ3 , (1.78)
and
σ3 + T0 = 0 if σ1 < −3σ3 , (1.79)
where the rock property T0 is called the uniaxial tensile strength.
In terms of centers and diameters of Mohr’s circles (i.e. in terms of σm and σd ,
respectively) these equations take the following forms:
σd2 − 16T0 σm = 0 if 4σm > σd , (1.80)
and
1
σm − σd + T0 = 0 if 4σm < σd . (1.81)
2
In these terms the Mohr’s circle equation (1.61) can be written as
1
f M (σm ) ≡ (σn − σm )2 + τt2 − σd2 = 0. (1.82)
4
The envelope of the Mohr circles, τt (σn ) satisfying criterion (1.80), can be found
from this criterion and two additional equations (see section 10.9 of Jaeger
et al., 2007):
∂ f M (σm )
f M (σm ) = 0, = 0. (1.83)
∂σm
This envelope provides a form analogous to equation (1.65) for expressing the
failure criterion for optimally oriented cracks:
τt2 = 4T0 (σn + T0 ). (1.84)
Using equation (1.81) for the Mohr envelope provides just its point (−T0 ; 0) on the
plane (σn ; τt ). This is a tip of parabola (1.84).
McClintock and Walsh (1962) included slip friction in the case of a crack closing
under normal compressive stress with the absolute value σc . For the case of a
compressive normal stress σn > 0 they derived a modified Griffith criterion (for
tensile stresses the original Griffith criterion expressed by the equation above is
still applied):

σ1 ( 1 + μ2f − μ f ) − σ3 ( 1 + μ2f + μ f ) = 4T0 1 + (σc /T0 ) − 2μ f σc . (1.85)
In terms of parameters of the Mohr circles this criterion has the following form:

σd 1 + μ2f − 2μ f σm = 4T0 1 + (σc /T0 ) − 2μ f σc . (1.86)
Using this equation and conditions (1.83) leads to a further modification of the
Mohr envelope (note that conditions τt > 0 and σn < σm must additionally be
taken into account):

τt = 2T0 (1 + σc /T0 ) + μ f (σn − σc ). (1.87)
For σc = 0 and T0 = Cc /2 this equation reduces to the Mohr–Coulomb criterion

(1.65). Note that this result is based on the sufficient condition of the unstable crack
growth. Therefore, the result of McClintock and Walsh (1962) can be considered
as a substantiation of the empiric Mohr–Coulomb criterion. To some extend this
clarifies the nature of the constants μ f and Cc (see Scholz (2002) and references
therein; equation 1.42 of Scholz (2002) is analogous to our equation above; note
also a misprint in the equation of Scholz (2002): an erroneous factor 2 in front of
the friction coefficient on its right-hand side).
1.2.4 Necessary conditions of crack growth and some results of

fracture mechanics
The mechanics of rock failure investigates conditions of their destabilization and
dynamics. This is a broad field of scientific research. Here we provide a brief
overview of some results that we use in this book. For a more comprehensive treat-
ment of the subject the reader should consult such books as Segall (2010), Jaeger
et al. (2007), Scholz (2002) and references therein.
We consider a plane penny-shaped crack of radius X/2 or a 2D crack infinite in
the x2 direction and of length X in the x1 direction. Griffith (1921, 1924) formulated
conditions for a growth of finite cracks based on an energy-balance consideration.
For a crack of length X to be in an equilibrium in a rock sample under a loading
stress τl (this stress must be specified correspondingly to the loading configuration)

the following energy balance must be fulfilled:
δ E s = δ E e + δ El . (1.88)
Here δ E s = γc δS is an energy increment needed to create a new crack surface δS

with the surface energy per unit area γc . For the crack to grow this energy increment
must be equal to the sum of energy contributions supplied by the rock sample itself
and by a system loading the sample. These are a loss of the sample’s elastic energy
δ E e and a loss of the energy of the loading system δ El , respectively. The loss of
the elastic energy δ E e occurs due to a decrease of the sample’s elastic stiffness G e
caused by the growth of the crack. The stiffness G e must be specified correspond-
ingly to the loading configuration and to the geometry of a crack. The loss of the
energy δ El of the loading system occurs due to the work of this system spent for
deforming the sample during the growth of the crack. Both of these energy changes
result from the work of the loading forces along the crack surface. Thus, together
they should be proportional to the length of the crack X , to its surface increment δS,
as well as to the density of the elastic energy in the sample, τl2 /(2G e ). Therefore,
the right-hand side of equation (1.88) must be proportional to X δSτl2 /G e .
If the energy balance (1.88) is disturbed so that its left-hand side is larger than
the right-hand side, then the crack will not receive sufficient energy to grow. On the
other hand, if the right-hand side of equation (1.88) is larger than the left-hand side,
then the crack will grow unstably and the sample will fail. The critical crack length
X = 2ac and the critical stress τl = τc are given by the balance condition (1.88):
Ge
γc = ac τc2 , (1.89)
Cg
where C g is a proportionality coefficient. For different modes of cracks, i.e. a ten-

sile crack (opening mode J = I), a longitudinal-shear crack (in-plane shear mode
J = II) and a transverse-shear crack (anti-plane shear mode J = III), and different
load configurations this coefficient is usually of the order of 1. For example (see
Jaeger et al., 2007, p. 309), for a tensile load of a thin sample with a tensile (2D)
crack under plane strain conditions, G e = 2μ(1 + ν) is the initial (pre-fractured)
Young’s modulus of the rock, and C g = π(1 − ν 2 )/2, with μ and ν being the
rock initial shear modulus and Poisson’s ratio, respectively. For the plane stress 2D
loading configuration, G e = 2μ(1 + ν) and C g = π/2. For a tensile load of a sam-
ple with a three-dimensional (3D) penny-shaped crack, G e is again the Young’s
modulus and C g = π(1 − ν 2 )/4. For a 2D transverse-shear crack under anti-plane
strain conditions, G e = μ and C g = π/4 (see Kanamori and Brodsky, 2004, and
references therein).
The left-hand side of equation (1.89) contains a combination of rock physical

properties only. The right-hand side of this equation contains a combination of the
crack length and the loading stress. For an arbitrary load τl and crack length X the
quantity

K J = τl C J X/2 (1.90)
is called the stress intensity factor. Here C J is a proportionality coefficient depend-
ing on the crack mode J . For example (Rice, 1980), C J = π for plane 2D cracks
(such cracks are infinite in their symmetry plane along the direction normal to the
propagation direction of their tips). However, this coefficient is more complex for
a circular crack.
From the linear elastic fracture mechanics (see Rice, 1980; and Jaeger
et al., 2007, pp. 311–314 and further references therein) it follows that on the plane
of the crack ahead of its tip the local stress perturbations produced by the fracture,
σ J , are given by
KJ
σJ = √ , (1.91)
2πr
where r is a distance from the crack tip.
Another result of this theory is that the crack-surface displacement u (absolute
value of the relativedisplacement of points across the crack) behind the crack tip
is proportional to τl (X 2 /4) − x 2 /G e , where |x| ≤ X/2 is the distance from the
crack center along the crack plane. For example, for modes I and II of 2D cracks,
u is given by the following expression (Rice, 1980; Scholz, 2002; Segall, 2010):
τl (1 − ν) 2
u = 2 (X /4) − x 2 , (1.92)
μ
and for mode III:
τl 2
u = 2 (X /4) − x 2 . (1.93)
μ
From the integration of a product of u taken in the vicinity of the propagating
crack tip with the stress (1.91), one obtains the energy increment ϒ released in
the process of the crack growth by the local stress forces per unit increment in the
crack length and per unit thickness in the x2 direction (see, for example, Jaeger
et al., 2007):
C g K J2
ϒ =2 . (1.94)
C J Ge
For the crack to grow this energy increment must be equal to, or exceed, the
energy increment required for creation of the new free surface:
ϒc = 2γc (1.95)
This will again give condition (1.89). This will also give a critical value for the
stress intensity factor:

K c = γc C J G e /C g . (1.96)
Sometimes K c is called the fracture toughness. For modes I and III values of K c
are from about 0.1 MPa × m1/2 for coal up to about 3.5 MPa × m1/2 for granite and
gabbro (see Scholz, 2002, and Jaeger et al., 2007, and references therein).
Griffith’s energetic criterion is a necessary rather than a sufficient condition of
unstable crack growth. On the other hand, the sufficient condition (we have consid-
ered it in the previous section) and the necessary condition are rather close. In other
words, creating tensile stresses of the order of critical ones simultaneously leads to
satisfying the thermodynamic Griffith criterion (see Scholz, 2002, pp. 6–7) and
vice versa.
The fact that under sufficiently high stress the atomic bonds will be broken and
the material in the vicinity of the crack tip must behave inelastically means that the
presence of the stress singularity in the results of the linear elastic fracture mechan-
ics of the type (1.91) is an artefact of the theory. There exist theoretical approaches
taking into account inelastic yielding at the crack tip (see Kanamori and Brod-
sky, 2004; Segall, 2010, and references therein). However, results of the linear
elastic fracture mechanics are applicable at distances r l y , where l y is the scale
of the region of inelastic deformations. In seismology the necessary criterion (1.95)
is often generalized to take into account the energy required for creating damaged
zones (Kanamori and Brodsky, 2004). This criterion does not take into account
seismic radiation occurring by dynamic propagation of the rupture. In this sense it
is a condition for a quasi-static growth of the crack. The criterion for dynamic crack
extension must also take into account the energy required for radiating seismic
wavefields depending on the rupture velocity (Rice, 1980; Kanamori and Brod-
sky, 2004). Moreover, recent studies (Bouchbinder et al., 2014) show the existence
of a length scale ln , where non-linear elastic effects are of importance. Scale ln is a
function of the crack velocity. It is shown that ln > l y . Thus, results of linear elastic
fracture mechanics are applicable for r ln .
1.2.5 Earthquake motions on faults

The considerations related to equations (1.88)–(1.95) assume that the loading stress
is equal to τl and the stress behind the tip of the crack is equal to 0. For earth-
quake mechanics, situations similar to the one considered by McClintock and
Walsh (1962) are especially interesting: the crack surface displacement leads not
to the complete stress release but rather to a decrease of the stress to the level τ f
supported by friction forces. Such situations are more relevant for the crack modes
II and III. In this case the stress τl must be replaced by the stress drop σ :
σ = τl − τ f . (1.97)
Linear elastic fracture mechanics shows that the mean relative displacement of
points across the crack u (given by the surface averaging of expressions like
(1.92) and (1.93)) has the following relation to the stress drop (Jaeger et al., 2007):
X σ
u = , (1.98)
2C μ
where C is a constant of the order of unity. For example, for a 2D plane vertical
strike-slip fault extending from depth X to the free surface, C = 2/π. For a circu-
lar of radius X/2 rupture surface, C = 3π(2 − ν)/(16(1 − ν)) = 7π/16, where
the last part assumes ν = 0.25.
The quantity u can be related to the strain distributed over a crack zone. Let
us firstly consider the following integral over a surface c closely surrounding the
crack:
1
ηij = (u i n j + u j n i )d 2 r, (1.99)
c 2
where n j (r) are components of the outward normal of c and u i (r) are displace-
ment components at a given point r of c .
In the case of a continuous elastic body with the surface c (i.e. a differentiable
displacement is given at all its points), Gauss’ theorem gives
1 ∂u i ∂u j 3
ηij = ( + )d r. (1.100)
V 2 ∂x j ∂ xi
The integrand here is the strain tensor, and V is the volume of rocks inside the
surface c . Thus, ηij /V is the volume-averaged strain.
In the case of a displacement discontinuity on a crack surface, integral (1.99)
still represents the volume-integrated strain in the crack zone. To show this one
considers a zero-thickness limit of the crack zone (see Rice, 1980). Let us denote
by + and − the sides of the crack. At the zero-thickness limit these sides will
geometrically coincide with each other and with the corresponding sides of the
surface c . The normals to the sides will point in opposite directions. Then:
1 + + 1 − −
ηij = (u i n j + u +j n i+ )d 2 r + (u i n j + u −j n i− )d 2 r
c+ 2 c− 2
1 + + 1 − +
= (u n + u +j n i+ )d 2 r − (u n + u −j n i+ )d 2 r
c+ 2 i j c− 2 i j
1
= (u i n +j + u j n i+ )d 2 r. (1.101)
c+ 2
Accordingly to conventional notation, the fault zone below a non-vertical fault (in
our notation, the rock contacting the c− -side of the fault) is called the foot wall.
The fault zone above the fault plane (in our notation it corresponds to the rock
contacting the c+ -side of the fault) is called the hanging wall. Thus, the slip vec-
tor (and more generally, the fault-displacement vector, i.e. the components u i )
describes the displacement of the hanging wall relative to the foot wall.
An equivalence between quantities (1.101) and (1.100) in the zero-thickness
limit of the crack zone results then from the following substitution (see Rice, 1980):
1 ∂u i ∂u j 1
ij ≡ ( + ) = (u i n +j + u j n i+ )δ D (c+ ), (1.102)
2 ∂x j ∂ xi 2
where δ D (c ) is the surface Dirac function turning a volume integral into a surface
integral over the discontinuity surface contained in the volume. This Dirac function
has physical dimension inverse to length.
Substituting (1.102) into (1.100) and performing the volume integration yields
ij d 3 r = pi j (r)d 2 r, (1.103)

V c+
where the quantity

1
pi j = (u i n +j + u j n i+ ) (1.104)
2
is called the potency density tensor (Segall, 2010).
Note that, according to (1.101), the right-hand part of equation (1.103) coincides
with ηij . Therefore, (1.99) and (1.100) become equivalent also in the case of a crack
zone. Therefore, the quantity
ij = pi j δ D (c+ ) (1.105)
represents the strain in the crack zone.
The potency density tensor can be rewritten in the following form:
1
pi j = u 0 (ûˆ i n +j + ûˆ j n i+ ), (1.106)
2
√
where u 0 = u i u i is the absolute value of the relative displacement of
points across the crack (also called the fault displacement). The unit vector with
components ûˆ i = u i /u 0 defines a direction of the fault-displacement vector.
If the fault-displacement vector represents just a slip along a planar crack surface
then the vector û will represent the slip direction of the c+ -side of the crack and
the vectors û and n+ will be orthogonal. Thus, ûˆ i n i+ = 0. Therefore, in the case of
a pure slip event the trace of the potency density tensor is equal to zero. Further, ûˆ i
and n i+ are components of unit vectors. Finally, we know that vector n+ points into
the upper half-space. These three conditions leave in the potency density tensor of
a slip event four independent components only. They can be expressed over the
δs
N
λs ϕs
Figure 1.7 Angles defining fault and slip orientations. The rock shown to the right
of the fault plane is the foot wall. The hanging wall is shown to the left of the fault
plane. Note that the rake has a negative value for this situation.
absolute value of the slip u 0 and the following three angles. The first one (the
strike φs ) is the azimuth of the straight line representing the intersection of the free
surface with the plane containing c+ . This angle is measured clockwise relative
to due north toward the dipping direction of the fault. In other words, the strike
direction is defined so that, for an observer looking into this direction, the fault
surface is dipping to the right (see Figure 1.7).
The strike angle can take values from the range 0 ≤ φs ≤ 2π. The second one
(the dip δs ) is the angle between the plane c+ and the free surface (0 ≤ δs ≤ π/2).
The third one (the rake λs ) is the angle measured from the strike direction of c+
to the direction of the slip vector. We follow here the notation accepted on pages
101–102 of Aki and Richards (2002) and assume that the rake takes values from
the range −π ≤ λs ≤ π . Negative angles are measured clockwise from the strike
direction. Note that for normal faults (such as the one shown in Figure 1.7) the rake
has negative values. For reverse faults (thrust faults) it is positive. If slip vector is
horizontal then one speaks about strike-slip faults.
Finally, in the case of a slip along the fault plane, the following useful relation
for the components of the potency density tensor follows from (1.106):

u 0 = 2 pi j pi j . (1.107)
Note that in a more general case of a fault displacement being not necessarily
parallel to the surface c+ the tensor of potency density has five independent com-
ponents. Thus, the four parameters mentioned before must be complemented by
one more, for example, the angle between the vector of normal to c+ and the slip
vector.
1.3 Elastic wavefields radiated by earthquakes

Growing cracks and slipping tectonic faults radiate elastic waves. A general
approach to formulating these wavefields was proposed by Backus and Mulc-
ahy (1976a,b). This approach was elegantly described by Rice (1980). Below we
follow his description, slightly modifying it for further applications in this book.
Let us consider a tectonic domain (or a sample of a rock) of volume V and sur-
face S with a fault (or a crack) inside. Elastic wavefields radiated by an earthquake
on this fault contribute to the strain field i j (r, t) of the tectonic domain. Dur-
ing the earthquake, part of the fault suffers a deformation iTj (r, t). One calls this
deformation a transformation strain. Earthquakes release tectonic stress. Thus, the
transformation strain is such that it will correspond to a zero-stress state (σi j = 0)
of a fault vicinity, if it is made free (isolated) from the rest of the tectonic domain.
Therefore, if a source region were isolated, it would be the case that:
i j (r, t)|isolated
sour ce = i j (r, t).
T
(1.108)
Using Hooke’s law (1.7) we can write the following formulation for such a
hypothetical zero-stress state:
0 = Ci jkl (i j (r, t) − iTj (r, t))|isolated
sour ce . (1.109)
This zero stress would be valid at any instant t because if the rupture starts to
propagate the stress will be released. Therefore, the total strain would be equal to
the transformation strain, as equation (1.108) states. Any additional strain would
vanish.
However, an earthquake-source region is a part of a tectonic domain. Then,
the right-hand part of equation (1.109) cannot be considered under conditions of
dynamic isolation. Thus, it must represent not a zero stress but rather an altered
stress state, where the strain is measured relative to the transformation strain:

σi j = Ci jkl kl (r, t) − klT (r, t) . (1.110)
Clearly, this stress state must satisfy the general elastodynamic equation (1.28):
∂ ∂ 2ui
Ci jkl kl (r, t) − klT (r, t) = ρ 2 . (1.111)
∂x j ∂t
Let us compare this equation with the general elastodynamic equation (1.28)
formulated for the case of a perturbation source in the form of a body-force density
vector with components Fi (r, t) and vanishing transformation strain:
∂ ∂ 2ui
(Ci jkl kl (r, t)) + Fi (r, t) = ρ 2 . (1.112)
∂x j ∂t
Thus, we conclude that, in the case of a perturbation source in the form of a trans-
formation strain, the effective volumetric force density is given by the following
expression:
∂
Fi (r, t) = − m i j (r, t), (1.113)
∂x j
where
m i j (r, t) = Ci jkl klT (r, t) (1.114)
is called the moment density tensor. Note that due to the symmetry of the tensor of
elastic stiffnesses the moment density tensor is also symmetric: m i j = m ji .
Therefore, equation (1.113) allows us to compute the displacement field pertur-
bation u i caused by an earthquake using results of the elasticity theory derived for
a perturbation source in the form of a body force. Thus, to obtain the displace-
ment one integrates a product of the effective force (1.113) and the elastodynamic
Green’s function G i j (r, r , t − t ) over the volume V and time. Assuming that the
altered stress (1.110) vanishes on the surface S of the tectonic domain and taking
into account the symmetry of the moment density tensor this integration is reduced
to the following result (Rice, 1980):
t
1
u i (r, t) = Hi jk (r, r , t − t )m jk (r , t )d 3 r dt , (1.115)
2 −∞ V
where the function

∂G i j (r, r , t − t ) ∂G ik (r, r , t − t )
Hi jk (r, r , t − t ) = + (1.116)
∂ xk ∂ x j
is a response of the medium to a perturbation source in the form of a singular
(δ-function-like) concentration of the moment density. Taking into account that
Green’s function G i j (r, r , t − t ) is a displacement response u i at the point r to
a singular body-force density acting in the direction x j at the location r , we see
that Hi jk (r, r , t − t ) is a displacement response u i at r to two couples of singular
body forces at r . Each of these force couples produces a response corresponding
to one of the two terms on the right-hand side of equation (1.116). The first term
corresponds to two forces acting in opposite directions along the axis x j . However,
they are applied at two differentially close points shifted in opposite directions
from the point r along the axis xk as shown in Figure 1.8. In the second term the
corresponding two forces act in opposite directions along the axis xk . They are
applied at two differentially close points spaced along the axis x j . In the case of
j = k such a combination of the four forces (i.e. of the two couples of forces)
xk
II
I
I r
II
xj
Figure 1.8 Double couple of forces corresponding to the Green function given by
equation (1.116).
is called the double couple. In the case of j = k the both force couples coincide
and represent a dipole of doubled forces (note, however, the factor 1/2 in front
of the right-hand-part integral in equation (1.115)). Such a combination of forces
is called the linear vector dipole. Note that all these combinations of two force
couples have resulting rotational (angular) momentum (in respect to the source
point r ) exactly equal to zero. Thus we conclude that, in a general case (including
also a generally anisotropic heterogeneous elastic medium), an earthquake radiates
seismic wavefields as a combination of double couples and linear vector dipoles
does. These double couples and linear vector dipoles can be distributed in a source
region V of the earthquake and in a time ts of the source process.
Using a standard approach of the linear system theory (e.g. forward and inverse
temporal Fourier transforms with multiplication and division by the factor iω),
equation (1.115) can be rewritten in an equivalent form with a response to a step
function of time instead of the impulse response:
1 t
∂m jk (r , t ) 3
u i (r, t) = E i jk (r, r , t − t ) d r dt , (1.117)
2 −∞ V ∂t
where
t
E i jk (r, r , t) = Hi jk (r, r , t )dt (1.118)
0
is the displacement response to a singular spatial distribution of double couples and
linear vector dipoles being step functions of time.
In the low-frequency range, where wavelengths of radiated wavefields are much
larger than the size of the source domain and at distances much larger than the
source domain, we can neglect the r -dependence of response E i jk (r, r , t) in
integral (1.117) and obtain:
1 t
∂ M jk (t )
u i (r, t) = E i jk (r, r , t − t ) dt , (1.119)
2 −∞ ∂t
where r represents an (averaged) source location and
Mi j (t) = m i j (r, t)d 3 r (1.120)

V
is a total moment tensor of the earthquake.
Equation (1.119) describes seismic wavefields radiated by a point earthquake
source. Therefore, in the low-frequency limit, it is reasonable to assume that the
duration time of the source process ts is very short. It is very short also in respect
to the wave propagation time from the source to observation points. Thus, in the
low-frequency limit we are interested in observational times such that t ts . The
moment tensor vanishes for times larger than ts . Then the integral (1.119) becomes
1
u i (r, t) = E i jk (r, r , t)M jk (ts ), (1.121)
2
or, in the spectral domain:
1
Ê i jk (r, r , ω)M jk (ts ).
û i (r, ω) = (1.122)
2
In other words, the low-frequency limit of the displacement spectrum is propor-
tional to the total moment accumulated at the end of the source process. The Fourier
transform of the response function E i jk is a factor in the spectrum of u i . It takes
into account various propagation effects of seismic wavefields in rocks. These are
effects like geometrical spreading, elastic anisotropy, reflection, refraction, diffrac-
tion and (more general) scattering on medium heterogeneities. There will also be
effects of inelastic attenuation and velocity dispersion often caused by fluids in
rocks, if the Green’s functions take these effects into account.
If we further consider the observation distances and the wavelengths to be sig-
nificantly larger than the source domain, but increase the frequency so that the
wavelengths become significantly shorter than the observation distances, then the
far-field approximation of the Green’s functions can be applied in equation (1.119).
In this approximation the elastic wavefields are equal to a sum of propagating
elementary waves radiated by a point source. In elastic homogeneous isotropic
media these are spherical P- and S-waves. The far-field Green’s function is given
by (1.39). The far-field displacements of the spherical P- and S-waves radiated by
a point-like earthquake will be correspondingly given by the following expressions
(see Rice, 1980):
1 ∂ M jk (t − r/c p )
u iP (r, t) = γi γ j γk , (1.123)
3
4πρc p r ∂t
1 ∂ M jk (t − r/cs )
u iS (r, t) = (δi j − γ i γ j )γk , (1.124)
4πρcs3r ∂t
where r = |r − r |, and the quantities γi = ri /r are the direction cosines.
1
0.5
0.1
0.05
0.01
0.005
0.001
0.1 0.5 1 5 10 50 100
ts
ω
2
Figure 1.9 Normalized amplitude spectrum (1.125).
Results (1.123) and (1.124) help to understand some general features of the fre-
quency spectra of earthquakes. Indeed, let us make a simple assumption that during
the process time ts the moment rate is constant and it is equal to zero outside of
this time interval. Then the moment rate is a boxcar function of length ts (i.e. the
moment itself is a ramp function) and of height M jk (ts )/ts . Its amplitude spectrum
(the absolute value of the forward Fourier transform of the boxcar function) is then
equal to

sin(ωts /2)

|M jk (ts )| . (1.125)
t ω/2
s
This is an even frequency function. Thus, we consider it for positive frequencies

only (see Figure 1.9). In the low-frequency limit this spectrum tends to a con-
stant equal to |M jk (ts )|. Such behavior corresponds to equation (1.122). In the
case of an isotropic homogeneous elastic medium the low-frequency limits of the
displacement spectra of the radiated P- and S-waves will be equal to
1
û iP (r, 0) = γi γ j γk M jk (ts ), (1.126)
4πρc3p r
1
û iS (r, 0) = (δi j − γi γ j )γk M jk (ts ). (1.127)
4πρcs3r
However, for frequencies higher than approximately
2
ωc ≡ , (1.128)
ts
the envelope of this spectrum starts to decay as 1/ω. The amplitude spectrum
(1.125) is proportional to an absolute value of a decreasing harmonic function.
It is equal to zero at the frequency values ω = 2π/ts , 4π/ts , 6π/ts . The quantity
ωc is called the corner frequency.
The results (1.123)–(1.125) describe wavefields radiated by a point source of
duration ts . If a rupture is finite in space then one can compute the radiated wave-
fields as a linear superposition of such point sources (at least approximately)
distributed along the rupture surface. Such a superposition takes into account a
finite rupture propagation velocity as well as directivity effects (similar to the
Doppler effect, e.g. shorter stronger pulses are radiated in the direction of the rup-
ture propagation and longer weaker pulses are radiated in the opposite direction).
To account for a finite rupture time td one must convolve the boxcar moment rate
function of duration ts with a boxcar function of duration td (see Shearer, 2009).
This is the so-called Haskell fault model (see Lay and Wallace, 1995). The corre-
sponding spectrum will be given by a product of two functions of the form (1.125)
with characteristic times ts and td . Thus, the high-frequency decay of the spectrum
will be given by 1/ω2 .
Similar spectra are actually observed. This fact gave rise to other similar models
of the earthquake spectrum. One of the most influential models was proposed by
Brune (1970):
|û(r, 0)|
|û(r, ω)| = , (1.129)
1 + (ω/ωc )2
which has the same main features as the Haskell pulse. Note that here the
corner frequency effectively corresponds to the intersection of the high- and low-
frequency asymptotes of the displacement spectrum (Aki and Richards, 2002). This
frequency (often modeled as an average over all radiation directions) is a func-
tion of the rupture geometry, size and velocity, and of the wave-signal speed. This
intuitive understanding is summarized in the following numerically and analyti-
cally supported relation between the rupture length X and the corner frequency
(Brune, 1970; Madariaga, 1976; Shearer, 2009):
4πkr cs
X= , (1.130)
ωc
where kr is a model-dependent and rupture-velocity-dependent numerical coeffi-

cient usually assumed to be of the order of 0.2–0.6.
Let us consider a simplified model of an earthquake as a slip with a surface-
average fault displacement u(t) along a plane rupture with the area A. Further we
assume that the rupture is located in a homogeneous isotropic elastic medium. Then
equations (1.102)–(1.106), (1.120), (1.114) and (1.25) give the following result for
the moment tensor:
Mi j (t) = Ci jkl pkl (t)δ D (c+ )d 3 x̂

V
1
= u(t)A(λδi j δkl + μ(δik δ jl + δil δ jk )) (ûˆ k n l+ + ûˆ l n +
(1.131)
k )
2
= M0 (t)(ûˆ i n +j + ûˆ j n i+ ),
where the quantity

M0 (t) = μu(t)A (1.132)
is called the scalar seismic moment. Combining this with equation (1.98) we obtain
2C uμ 8C M0 C 0 M0
σ = = = , (1.133)
X πX 3 X3
where M0 = M0 (ts ) = μu(ts )A = μuπ X 2 /4, and X stands for a characteristic
length of the rupture. In the case of a circular rupture, X is the diameter of the
rupture. Geometry-controlled constants C0 = 8C /π and C are of the order 1.
Further, using equation (1.130) to estimate the characteristic length of the rup-
ture from the corner frequency, one obtains an equation often used to estimate the
stress drop:
3
C ωc
σ = M0 , (1.134)
π 2πkr cs
where, for example, C = 7π/16 for a circular rupture. Note that this equation
is based on a series of critical assumptions about the rupture processes that may
not be necessarily valid (see Beresnev, 2001). Moreover, this stress-drop estimate
is very sensitive to the corner frequency, which is in turn a parameter influenced
by many not completely determinable factors. These are, for example, the seismic
attenuation, local site effects, heterogeneity of the medium and some instrumental
features.
Finally, the scalar moment is related to the moment magnitude Mw by the
following definition (Kanamori, 1977; Shearer 2009):
2
Mw = (lg M0 − 9.1), (1.135)
3
where the moment is taken in Nm (SI system). Note also that, in this book, we use
the International Standard notation lg ≡ log10 and ln ≡ loge . We imply moment
magnitudes when addressing earthquake magnitudes throughout this book.
Multicomponent records of P- and S-waveforms can be used to find focal mecha-
nisms of earthquakes. Such solutions usually represent fault plane (double-couple)
approximations of tectonic motions at earthquake faults. Moreover, more complete
moment-tensor solutions can be found. They also include non-double-couple com-

ponents of the moment tensor, i.e. isotropic and compensated-linear-vector-dipole
(CLVD) ones (see Shearer, 2009, section 9.2).
1.4 Introduction to microseismic monitoring

Microseismic monitoring is a method of seismic investigation of the under-
ground. It is based on detection and location of small-magnitude earthquakes
(usually called microearthquakes or sometimes also called seismic emission)
often occurring in rocks due to various artificial or natural processes. Microseis-
mic monitoring is a powerful method of geophysical reservoir characterization.
It can contribute to delineation and imaging of reservoirs and their heteroge-
neous structures, characterization of hydraulic and elastic properties of rocks,
understanding of geomechanical features like tectonic stresses and pore-pressure
distributions.
Results of microseismic monitoring might be further combined with four-
dimensional (4D) reflection seismics, reservoir simulations, borehole measure-
ments and stimulation and production data. Such integrated information provides
a rather complete picture of relevant processes in underground targets. One of the
most common applications of microseismic monitoring is the mapping of hydraulic
fractures in hydrocarbon reservoirs (Urbancic and Baig, 2013). Maxwell (2014)
gives a detailed description of many practical, historical, technical and environ-
mental aspects of this application.
1.4.1 Detection of seismic events

To record seismic events, arrays of seismic receivers are required. Seismic receivers
(also called geophones) can be placed on the Earth’s surface (see Duncan and Eis-
ner, 2010) and/or in several or in a single monitoring borehole. A schematic sketch
of seismic monitoring is shown in Figure 1.10. Figures 1.11 and 1.12 show an
example of real borehole-monitoring system for three hydraulic fracture stages at
a shale-gas reservoir.
Multicomponent seismic receivers record components of vectorial seismic wave-
fields (i.e. components of motions of rock particles) as functions of time (i.e.
seismic traces or seismograms; see Figures 1.13 and 1.14). Application of arrays
of seismic receivers and their ability to record different components of seismic
wavefields allow for detection of seismic waves radiated by earthquakes. This is
usually done by different types of analysis of seismic traces, such as their energy
analysis including ratios of energy averaged in short and long time windows of
seismic traces, their statistical analysis, their integral transforms including Fourier
A B
B
C
Figure 1.10 A sketch of possible acquisition geometry of microseismic moni-

toring. Multicomponent geophones can be placed in several boreholes or in a
single monitoring borehole. Multicomponent geophones also can be placed on
the Earth’s surface. Geophones (denoted by the cubes) record seismic wavefields
radiated by a microearthquake (a spot to the right). A sketch of the triangulation
principle of event location using travel time information at geophones A, B and C
(the elements of differential isochrones are schematically shown by corresponding
arcs) is given on the left.
and wavelet analysis, analysis of particle motions, or combinations of those (see,

for example, the literature review in Rentsch et al., 2007). Sometimes one also
applies the technique of master event (a matched-filter technique) originally pro-
posed for the detection of global seismic events (Shearer, 1994). The data basis
of template earthquakes is enriched with time and additional events can be also
detected retrospectively.
Further processing of microseismic wavefields usually requires identification of
seismic phases (for example arrivals of the P- and S-waves) and the picking of their
arrival times. In spite of the fact that the picking of arrival times can be accom-
plished with automatic algorithms, manual picking is still frequently performed to
increase the accuracy of the travel-time information.
1.4.2 Seismic multiplets

The waveform similarity is important for event location. Seismic events charac-
terized by a high similarity are referred to as a seismic multiplet. Figure 1.15
shows an example of seismic traces of a multiplet composed of three different
microearthquakes at Cotton Valley.
Such events probably occur on the same structures and have the same tec-
tonic mechanisms. To quantify the waveform similarity one can use their
X
Z Y
Figure 1.11 Hypocenters (dots) of microseismic events induced by three

hydraulic-fracturing stages conducted from a horizontal section of a borehole
(black line) in a shale-gas reservoir in Canada. The corresponding perforations are
shown as small spheres. The distances between the perforations are approximately
100 m. The depth range of the microseismic clouds is approximately 1700–
1800 m. The observation system was composed of two chains of three-component
geophones (small cubes) located in two different monitoring boreholes (two other
black lines with cubes). The asymmetry of the microseismic clouds in respect to
the perforation locations is due to the asymmetric position of geophones in respect
to the perforated borehole (see also Figure 1.12). The figure is courtesy of Anton
Reshetnikov, Freie Universität Berlin.
cross-correlation coefficients. For example, Scholze et al., (2010) used the fol-
lowing definitions. The normalized cross-correlation function ci j (t) between two
traces u i (t) and u j (t) is defined as

u i (t ) u j (t − t) dt
ci j (t) = 12 . (1.136)
2 2
u i (t )dt × u j (t )dt

The cross-correlation coefficient is then defined as the absolute maximum of the

cross-correlation function, CCi j = max[|ci j (t)|]. It can be calculated for the com-
plete seismograms or separately for specified time windows containing selected
wave phases (e.g. P- or S-waveforms).
Clustered multiplet events can be observed in the seismicity induced by artificial
rock stimulations as well as in natural seismicity. Multiplet clusters can frequently
Figure 1.12 The microseismic clouds, the monitoring system and the boreholes
shown in Figure 1.11 but now in the map view. The figure is courtesy of Anton
Reshetnikov, Freie Universität Berlin.
40
35
30
25
# Trace
20
15
10
0
0 1 2 3
Time [s]
Figure 1.13 Seismic traces of the vertical and two horizontal components of the
wavefield radiated by a microearthquake that occurred during a water-injection
experiment in the pilot borehole of the German Continental Deep Drilling Project
(KTB). The geophone array was composed of different types of seismic receivers.
Five three-component 4.5 Hz seismic sensors were deployed around the KTB in
shallow boreholes (25–50 m depth). Additionally, eight three-component 1 Hz
sensors were deployed on the surface. The sampling rate of these 13 shallow geo-
phones was 200 Hz. One more three-component 15 Hz geophone was installed
at 3500 m depth in the main KTB borehole. Its sampling rate was 1000 Hz. Cor-
responding seismic traces of the vertical and two horizontal components have
numbers 1, 2 and 3, respectively. The figure is courtesy of Jörn Kummerow, Freie
Universität Berlin. (From Shapiro, 2008, EAGE Publications bv.)
50 60 70 80 50 60 70 80 50 60 70 80
0.2 0.2 0.2
0.3 0.3 0.3
0.4 0.4 0.4
0.5 0.5 0.5

Time (s)
0.6 0.6 0.6
0.7 0.7 0.7
0.8 0.8 0.8
0.9 0.9 0.9
X component Y component Z component
Figure 1.14 Seismograms of east, south and depth components (from the left,
respectively) of the wavefield radiated by a tectonic microearthquake at the San
Andreas Fault. The seismograms were recorded by a borehole array of the Pauls-
son Geophysical Services Inc. (P/GSI) in the main borehole of the San Andreas
Fault Observatory at Depth. The array is composed of 80 three-component 15 Hz
borehole geophones located at depths between 878 m and 1703 m below sea level.
The horizontal axis represents the number of geophones (the traces shown were
recorded by geophones with numbers 43–80). The vertical axis is time in seconds,
and the sampling rate is 0.25 ms. (Modified after Reshetnikov et al., 2010.)
Figure 1.15 A multiplet of three different microseismic events recorded at the

same borehole geophone during a hydraulic fracturing at Cotton Valley. The traces
are 0.1 s long. The arrivals of the P- and S-waves are clearly seen at the top and
in the middle of traces, respectively. The figure is courtesy of Karsten Stürmer,
Freie Universität Berlin.
be located with high precision. Based on the processing of multiplet events,

Kummerow (2013) proposed an automatic approach of enhancing consistency in
determining arrival times of direct seismic P- and S-waves. He cross-correlated
microseismic borehole array data for such events. This leads to a cross-linking of
waveforms of different events in a multiplet recorded by several closely spaced
receivers. The picked travel times are then iteratively corrected by forcing a con-
sistent time alignment of such waveforms. Improved travel times lead finally to
better event locations.
Following the work of Lin and Shearer (2007) on tectonic earthquakes,
Kummerow et al., (2012) proposed applying microseismic multiplets to estimate
the spatial distribution of the velocity ratio of the longitudinal and shear waves,
c p /cs . This ratio is directly related to the Poisson ratio. High-precision P and
S arrival time differences must be calculated for pairs of microseismic events.
Multiplets are highly suitable for this task. Then, the velocity ratios are determined
from the slope of distributions of differential P- versus S-travel times. If done for all
elementary spatial cells, this will provide a 3D c p /cs -ratio image. Such a procedure
requires approximate event locations as an input.
1.4.3 Location of seismic events

Travel-time characteristics of P- and S-waves (e.g. their arrival times, curvatures
of arrival time surfaces as functions of receiver locations, and differences in
arrival times of longitudinal and shear waves) as well as their vectors of particle
displacements (i.e. their polarizations) can be used for locating earthquake
hypocenters. A hypocenter of an earthquake is the starting point of the correspond-
ing rupture. To locate an earthquake means finding the spatial coordinates of its
hypocenter and its origin time. It is one of the most important issues in earth-
quake seismology (Thurber and Rabinowitz, 2000) as well as in the microseismic
monitoring (Rentsch et al., 2007).
Most location procedures require P- and S-wave arrival times as well as a veloc-
ity model between the hypocenter and the receiver array. In some approaches,
expected arrival times are calculated and compared with measured ones for every
receiver. Then, event locations and sometimes velocity-model updates are found
by grid-search-type optimization algorithms which minimize arrival-time residu-
als (Lomax et al., 2009). Often, different simplified variations of such approaches
are applied. For example, for each geophone of the monitoring array one uses
differences between P- and S-wave arrival times and a given velocity model to
calculate spatial surfaces (differential isochrones) of possible hypocenter loca-
tions. A hypocenter is then assigned to an intersecting region of these surfaces
(of course, they are spheres, in the case of simple constant isotropic velocity
models). Other location approaches use full waveform seismograms. These are
migration-type location methods (Rentsch et al., 2010) or methods using seis-
mogram cross-correlations in combinations with arrival-time-based wavefield
characteristics (Richards et al., 2006).
Seismic multiplets and seismogram cross-correlations are extensively used in
relative earthquake location methods such as the double-difference method of
Waldhauser and Ellsworth (2000). In such approaches the cross-correlation of P-
and S- wave traces are used for precise estimates of travel-time differences for pairs
of earthquakes. Then residuals between observed travel-time differences and the
velocity-model-based predictions of these differences are minimized. This leads
to high-resolution images inside seismicity clusters. Some event-location proce-
dures exploit the dependence of waveform similarity on event spacing. Given an
initial set of precise absolute earthquake locations, the maximum waveform cross-
correlation coefficients for all possible pairs of events are calculated, and a relation
between event separation and cross-correlation coefficients is established. This
is used to locate further events by calculating their waveform similarity relative
to all located events. An advantage of this approach is that it can be applied to
events registered even by a single seismic sensor. Kummerow (2010) proposed this
method and applied it to microseismic data recorded during a hydraulic experi-
ment at the Continental Deep Drilling Site (KTB) in Germany. The number of
reliably determined hypocenters was increased by a factor of about eight com-
pared with standard location methods. Another variation of this approach inverts
both the measured arrival times and cross-correlation values of the waveforms
for the hypocenter coordinates. Such algorithms have the potential to signifi-
cantly better locate seismic events in comparison to the purely arrival-time-based
approaches.
An important aspect of event location is a construction of the velocity model
(Pavlis, 1986). The elastic anisotropy of rocks must be accounted for. Thus the
velocity model must also include parameters characterizing the anisotropy. Elastic
moduli Ci jkl or Si jkl can be used for this task. Thomsen (1986) proposed a conve-
nient way to describe seismic anisotropy using the so-called anisotropy parameters.
In many relevant situations of microseismic monitoring in sedimentary rocks the
transverse isotropy is a good approximation. For the transverse isotropy with the
vertical (x3 ) rotational symmetry axis, Thomsen (1986) introduced the following
three anisotropy parameters:
c11 − c33
a ≡ , (1.137)
2c33
c66 − c55
γa ≡ , (1.138)
2c55
(c13 + c55 )2 − (c33 − c55 )2

δa ≡ , (1.139)
2c33 (c33 − c55 )
where, for elastic moduli, we used the contracted notation discussed in

Section 1.1.4. We also added the index a to Thomsen’s original notation for the
anisotropy parameters in order to distinguish them from other notations used in
the book. To define the velocity model for a homogeneous transversely isotropic
medium with the vertical symmetry axis (the so-called VTI medium), these
three parameters are usually complemented by the velocities of P- and S-waves
propagating along the x3 direction:
c33
c p0 = , (1.140)
ρ
c55
cs0 = . (1.141)
ρ
These two velocities and the three Thomsen parameters can be used to recon-
struct five independent elastic moduli of a transverse isotropic medium. For such
a reconstruction the sign of (c13 + c55 ) must be additionally known. Usually it
is assumed that c13 + c55 > 0 (Tsvankin, 2005). In Section 1.1.4 we saw that,
in isotropic media, c33 = c22 = c11 = λ + 2μ, c23 = c13 = c12 = λ and
c66 = c55 = c44 = μ. Therefore, in isotropic rocks the anisotropy parameters
vanish. This indicates that the anisotropy parameters can be considered as a mea-
sure of the strength of anisotropy. Owing to the stability conditions (see Section
1.1.4) the anisotropy parameters cannot be less than −0.5 (Grechka, 2009). They
have no upper bounds. Shale is often characterized by a very strong anisotropy,
with the anisotropy parameters above 0.3. Possible values of a above 0.5 and
the typical relation a > δa have been observed (Tsvankin, 2005). The case for
Thomsen parameters smaller than 0.1 is referred to as weak anisotropy. Sand-
stones are often weakly anisotropic. Carbonates are frequently nearly isotropic.
Crystalline rocks like basalt and granite often do not show significant elastic
anisotropy. However, fractures, an oriented texture and tectonic stress can change
this situation.
In Section 1.1.5 we saw that, in elastically anisotropic rocks, three different types
of seismic body waves can be observed. This is often the case for the microseismic
wavefieds in shale (see Figure 1.16).
Even in the case of weak seismic anisotropy, significant location errors can occur
by neglecting this effect. Some inversion-based approaches to event location and
macromodel construction have been proposed. For example, Grechka and Yaske-
vich (2013) proposed a simultaneous inversion and location approach for general
Vertical North East
0.45
0.50
0.55
Time (s)
0.60
0.65
Figure 1.16 Traces of a microseismic event induced by the hydraulic fracturing

in a shale-gas reservoir in Canada. The observation system is shown in Figures
1.11 and 1.12. Approximate directions of polarization components are indicated
by the titles of the seismograms. Arrivals of the quasi P-wave are clearly recorded
by the all 30 geophones. The arrivals are approximately seen in the time inter-
val 0.48–0.51 s on the first nine geophones and as a curved moveout in the time
interval 0.54–0.56 s on the last 21 geophones. The slower quasi S-wave (which
corresponds to the qSV-wave in a VTI medium) is seen on all the last 21 geo-
phones as a moveout in the time interval 0.63–0.65 s. The quicker quasi S-wave
(which corresponds to the SH-wave in a VTI medium) is seen on the north- and
east components of the last 21 geophones. Its first arrivals are approximately in
the time interval 0.60–0.62 s. Both these qS-waves can be also seen on the first
nine geophones. Their arrivals partially interfere and closely follow the qP-wave.
A moveout of the slower qS-wave reflected by a layer interface is seen on the
vertical component of the sixth to twelfth geophones (from the left) of the second
array in the time interval 0.65–0.67 s. The complexity of the waveforms is due to
an approximate tilted VTI symmetry of the medium, its heterogeneity, the related
ray bending and not-exact geophone orientations. At this site and depth interval
of this layered shale deposit Yu (2013) estimated values of Thomsen parame-
ters in the range 0.15–0.4 for a , in the range 0.15–0.2 for δa and in the range
0.15–0.7 for γa . The figure is courtesy of Anton Reshetnikov, Freie Universität
Berlin.
triclinic anisotropic media. Yu et al. (2013) proposed a variation of this approach

for transversely isotropic layered media. Velocity anisotropy is taken into account
in an additional step after conventional event locating. In this step, the isotropic
velocity model and corresponding locations serve as an initial model. Then, the
anisotropy parameters as well as the resulting locations are obtained simultane-
ously. The algorithm uses ray theory for travel-time computations and known
numerical approaches for non-linear optimization.
0 500 1000 0 500 1000
1000 1000
1500 1500
2000 2000
2500 2500
3000 3000
Figure 1.17 On the left: a result of microseismic reflection imaging at the main
borehole of the San Andreas Fault Observatory at Depth. On the right: a result
of reflection seismic imaging from the free surface at the same location. The
borehole is shown by the black line. The scale is given in meters. Both images
are similar; however, the microseismic image has significantly higher resolution.
(Modified after Reshetnikov et al., 2010.)
1.4.4 Microseismic reflection imaging

An interesting possibility of using microseismic information is given by treating
microearthquakes as active sources. First, high-precision location of microseismic
events is necessary. Second, one migrates the corresponding reflected wavefield
using one of reflection seismic-imaging techniques. For example, Reshetnikov
et al. (2010) used a Fresnel-volume-limited Kirchhoff summation (a directional
migration algorithm) to obtain a high-resolution image of a subsurface region
illuminated by microseismic wavefields. They applied the method to several micro-
seismic events recorded by a borehole array in the SAFOD (San Andreas Fault
Observatory at Depth) main hole and compared the findings with results of surface
seismic reflection profiling (Figure 1.17).
Usually, the number of located microseismic events is much larger than the
number of sensors. Using the principle of reciprocity, one can swap sources and
receivers. Thus, by such a type of imaging one considers a monitoring system as
several “sources” and a cloud of microearthquakes as an array of spatially dis-
tributed “receivers.” In the case of the Basel geothermal experiment it was possible
to obtain a high-resolution reflection image (Reshetnikov et al., 2013). More-
over, it seems that identification of singular large-scale fractures was possible. The
consistency and extremely high resolution of the method can be enhanced in its
multisource multireceiver configuration.
Frequently we observe significant microseismic reflections, indicating the
presence of strong heterogeneities. Numerical and theoretical studies of waves
reflected at thin fluid layers (see Oelke et al., 2013) show that in the microseismic
frequency range ( f = 50–400 Hz) hydraulic fractures and thin fluid-filled cracks
(modeled as fluid layers of thicknesses 10−3 –10−2 m) can produce strong reflection
coefficients (0.2 and higher). In addition, the double-couple sources corresponding
to microseismic events can produce rather complex reflection signatures due to
non-radially symmetric radiation patterns.
2
Fundamentals of poroelasticity
Fluid-saturated rocks are multiphase media. Elastic stresses are supported there by
the solid phase of rocks. Pore pressure is supported by saturating fluids. In such
media the elastic stresses and the pore pressure are coupled.
The mechanics of poroelastic media was developed by Maurice Biot in
a series of his seminal publications in the 1940s–1970s (see, for example,
Biot 1956, 1962). This theory was further developed and reformulated by several
scientists. A very incomplete list of corresponding works includes the publica-
tions of Gassmann (1951), Brown and Korringa (1975), Rice and Cleary (1976),
Chandler and Johnson (1981), Rudnicki (1986), Zimmerman et al. (1986), Detour-
nay and Cheng (1993) and Cheng (1997).
We consider induced microseismicity as a poromechanical phenomenon. Coussy
stated in the preface to his book Poromechanics (2004): “We define Poromechanics
as the study of porous materials whose mechanical behavior is significantly influ-
enced by the pore fluid.” In this book we use several approximations derived from
the theory of poroelasticity, which is part of poromechanics.
We start our introduction to poroelasticity by considering elastic compliances
of porous rocks. Here we follow and correspondingly modify the approach of
Zimmerman et al. (1986) and Brown and Korringa (1975). Following Detournay
and Cheng (1993) we consider a very small elementary sample of a porous fluid-
saturated rock with a characteristic length dl, which is assumed to be more than
ten times larger than the typical size of pores and grains. Such a sample can be
also considered as a representative volume of a statistically homogeneous porous
medium. Both solid and fluid phases are assumed to be connected in a 3D rock
structure. The pore space of a rock sample is defined as the total space occupied by
all connected voids (including fractures and pores) in the sample. Let us denote by
V p the volume of the pore space of a sample of volume V . The ratio
Vp
φ= (2.1)
V
48
n
Σc
Ψp
Figure 2.1 A sketch of external and internal surfaces of a porous system.
is the so-called connected porosity of the rock. Throughout the book we consider
total porosity to mean connected porosity. Isolated pores are understood as part of
the solid material (grain material).
Such a porous sample has two closed surfaces on the sample scale: its external
surface c and surface p of its pore space (see Figure 2.1). The surface p of the
connected pore space is the sample’s internal surface.
2.1 Linear stress–strain relations in poroelastic materials

We consider a thought experiment where a uniform confining stress, σi j (also called
total stress), is applied to the external surface c of the rock sample. Simultane-
ously, a pore pressure, Pp , is applied to the internal surface p . Surface c is the
external surface of the solid part of the rock. Simultaneously it also seals pores. At
a given point r of the surface c the applied traction τ is then
τi = σi j n j (r), (2.2)
where n j (x̂) are components of the outward unit normal to c .
Let us assume that the confining stress and/or pore pressure has changed from
the load state (σi0j , Pp0 ) to the load state (σi j , Pp ). As a result, points of the external
surface have been displaced by u i (r). The displacement is assumed to be very small
in comparison with the size of the rock sample.
Following Brown and Korringa (1975), we introduce a symmetric tensor (note
also the analogy to equation (1.99)):
1
ηi j = (u i n j + u j n i )d 2 r. (2.3)
c 2
50 Fundamentals of poroelasticity
In the case of a continuous elastic body replacing the porous rock (i.e. a
differentiable displacement u is given at all its points) the Gauss’ theorem gives

1 ∂u i ∂u j
ηi j = + d 3 r. (2.4)
V 2 ∂x j ∂ xi
The integrand here is the strain tensor and V is the volume of the sample. Thus,
i j = ηi j /V is the volume averaged strain.
We introduce also another symmetric tensor,
1
ζi j = (u i n j + u j n i )d 2 r, (2.5)
p 2
where p is the surface of the pore space, r is a point on this surface, u i is a

component of the displacement of points r on this surface slightly deformed by
changing the load, and n i is a component of the outward normal to this surface
(the normal is directed into the space of pores, see Figure 2.1). At points where
surface c seals the pores it coincides with surface p ; however, their normals
have opposite directions.
If we assume that the pore space is filled with some material (e.g. a fluid, clay or
cement) then, analogously with tensor ηi j , a quantity −ζi j will denote a volume-
averaged strain of this material multiplied by its volume. Note that the minus sign
is due to the direction of the normal of surface p . Thus, the quantity −ζii denotes
a change of the volume of this pore-filling material.
In the linearized theory of poroelasticity the initial load state is not of
importance. It is especially so if incremental small-strain effects are considered. For
simplicity we will assume that the initial load state corresponds to σi0j = Pp0 = 0.
Later, in Section 2.9, we consider some effects of the initial stress state of rocks.
In realistic experiments a uniformly distributed scalar pore pressure Pp can be
applied to surface p . However, on the following pages (until equation (2.19)) we
will assume a more general, albeit abstract, possibility that a uniformly distributed
f
stress tensor σi j is applied to the surface p (note that the stress compressional
with respect to the solid phase is negative). In spite of its quite artificial character,
this assumption can be advantageous. For example, it will allow us to imagine a
load configuration where the compliance tensor of the grain material can be mea-
sured (at least in an average sense) without destroing the porous sample (see our
discussion of experiments with an unjacketed sample below). If the load on surface
f
p is hydrostatic, then σi j = −Pp δi j . The effective stress is defined as:
f
σiej = σi j − σi j . (2.6)
Note that in this equation we introduce only a notation, “effective stress.” At

this point we do not discuss any physical quantity for which this stress is indeed
“effective.” We will discuss this later in this chapter and especially in Section 2.9.6.
Further, for a completely hydrostatic load (i.e., also the confining stress is given
by a confining pressure σi j = −Pc δi j ) we have:
f
σiej = σi j − σi j = (−Pc + Pp )δi j = −(Pc − Pp )δi j . (2.7)
By analogy with Brown and Korringa (1975) and Zimmerman et al. (1986)
we introduce different compliances of an anisotropic porous body correspond-
ing to different configurations of its loading. Indeed, it is possible to apply loads
to both surfaces of a porous sample (see Figure 2.2). It is equally possible to
measure displacements of these two surfaces by loading any one of them. Thus,
there are four different combinations of loading a surface and measuring displace-
ments of the same or of the other surface. We assume that the applied incremental
load is small (the sample can be pre-stressed; in this case we discuss incremental
stresses and strains). Equally we assume that the resulting incremental displace-
ments are small (so that the strains of the sample and of its pore space are small).
Such loading experiments lead to formulating four Hooke’s laws relating the cor-
responding strain and stress tensors. This allows us to define four compliance
tensors.
Figure 2.2 A jacketed porous sample (the doubled boundary of the sample sym-
bolizes its jacketting envelope) provides a possibility to apply independent loads
to its external and internal (pore-space) surfaces. Thin curved arrows inside
the sample symbolize uniformally distributed loading of pore-space surface p .
Thick arrows outside the sample are applied to its external surface c . Note that
in the case of a fluid-saturated sample this figure is a draft of a realistic laboratory
experiment.
Figure 2.3 A sketch of a loading configuration, where a jacketed sample provides

the possibility of establishing drained conditions.
Generally surfaces c and p are isolated from each other and can be subjected
to different loads. Let us first assume that only the surface c has been subjected
f
to a changing load. Thus, the load σi j has been unchanged. In the case of fluid
saturation this is a load configuration of a jacketed sample organized in such a way
that the fluid can freely drain away from the sample (see Figure 2.3). The so-called
drained compliance tensor can be then introduced analogously to Hooke’s law for
elastic media:
1
ηi j |σ f = Sijkl σkl .
dr e
(2.8)
V
The compliance tensor introduced here characterizes the elastic properties of the
drained skeleton of the rock. In equation (2.8), and later in other similar equa-
tions expressing different Hooke’s laws for poroelastic systems, corresponding
dr
components of compliance tensors (here Sijkl ) are defined as partial derivatives of
strain components (here ηi j /V ) over stress components (here σkle ), respectively
(see Brown and Korringa, 1975; and Shapiro and Kaselow, 2005). Note that in
such partial derivatives the symmetric off-diagonal stress components (σkle and σlke )
are formally assumed to be independent, in spite of the fact that in reality they
are equal.
Another loading configuration (using an unjacketed sample) can be defined in
the following way. Both surfaces c and p are loaded by the same changing part
of the stress (see Figure 2.4). Therefore, the effective stress remains unchanged.
Such a homogeneous distribution of the stress on the internal and external surfaces
of the sample is equivalent to a distribution of the stress in a homogeneous elastic
Figure 2.4 An unjacketed sample providing a possibility to establish constant-

effective-stress conditions. For a changing part of the load (symbolized by thin
arrows) we assume an equivalent access to the internal and the external surfaces
of the sample. In the case of a fluid-saturated sample this can be achieved by
filling the space around the sample by the saturating fluid. The changing part of
the load is then the fluid pressure.
medium constructed from the same material as the skeleton of the rock. Thus,
the following form of the Hooke’s law introduces the tensor of compliance of the
skeleton material. The skeleton material is also often called the grain material. In
the case of a heterogeneous grain material the corresponding compliance tensor
characterizes its elastic properties in an effective (averaged) sense:
1 gr f
ηi j |σ e = Sijkl σkl . (2.9)
V
The two compliance tensors introduced above can be found from measurements
of displacements of the external surface c . Measurements of displacements of
the internal surface p provide two more compliance tensors. The first one can
be called the compliance tensor of the pore space. It is obtained from the above-
described experiment with an unjacketed sample (under a constant effective stress,
see Figure 2.4):
1 p f
− ζi j |σ e = Sijkl σkl . (2.10)
Vp
If the material of the solid skeleton of the rock (i.e. the grain material) is homo-
geneous and linear, then the condition of a constant, uniformly distributed loading
stress will be equivalent to the replacement of the material in pores by the grain
material. Thus, the porosity does not change. Moreover, in this case the volume-
averaged strain is independent of the averaging domain. This yields the equivalence
of equations (2.9) and (2.10) and leads to the following statement:
p gr
Sijkl = Sijkl . (2.11)
Finally, one more (but not-independent) compliance can be estimated from dis-
placement measurements of the internal surface. If we assume that effective stress
changes but the stress σ f remains constant, then:
1
− ζi j |σ f = Sijkl σkle . (2.12)
V
Using the reciprocity theorem analogously to Brown and Korringa (1975) we
obtain (see the Appendix to this chapter):
gr
Sijkl = Sklij
dr
− Sklij . (2.13)
Let us consider a strain i j = ηi j /V of a porous elastic sample due to a loading
by changing both stresses, σ e and σ f . Note that we assume the strain to be small
and neglect all non-linear terms (in respect to any displacement gradients). Under
arbitrary (but small) changes of a load the strain will change due to applying incre-
mental σ e by keeping a constant stress σ f plus an effect of applying incremental
σ f from inside and outside (i.e. keeping σ e = const.). This is a consequence of a
linear superposition of deformation effects of independent loads. Therefore, using
equations (2.8) and (2.9) as well as the definition of the effective stress (2.6) we
obtain
f gr f
i j = Sijkl
dr
(σkl − σkl ) + Sijkl σkl . (2.14)
Analogously we can consider the strain of the pore space (later we will explicitly
assume that the pore space is indeed saturated by a fluid). This strain will change
due to σ e by keeping a constant stress σ f plus an effect of applying incremental σ f
from inside and outside of the porous sample (i.e. keeping σ e = const.). Therefore,
using equations (2.10), (2.12) and (2.13) we obtain:
1 p f 1 dr gr

f

− ζi j = Sijkl σkl + Sklij − Sklij σkl − σkl . (2.15)
Vp φ
Let us assume that the pore space is saturated by a fluid or filled with another
f
solid material. The following strain of the infilling material i j results just from the
f
action of a homogeneously distributed stress σkl in it:
f f f
i j = Sijkl σkl . (2.16)
In the case of an undrained system, when the infilling material (including a fluid)
has no possibility to migrate (at least on average) outside of the solid frame (i.e.
the mass of the material infilling the pore space is constant), two strains (2.15)
and (2.16) must be equal (see also the comments after equation (2.19)), and thus:
f f p f 1 dr gr

f

Sijkl σkl = Sijkl σkl + Sklij − Sklij σkl − σkl . (2.17)
φ
In this case the so-called undrained compressibility of the saturated porous rock
u
Sijkl can be calculated. For this the strain (2.14) must be rewritten by explicitly
using the undrained compressibility and the confining stress:
f gr f
u
Sijkl σkl = Sijkl
dr
(σkl − σkl ) + Sijkl σkl . (2.18)
f
Expressing tensor σkl from equation (2.17), substituting it into equation (2.18) and
eliminating from there tensor σkl we obtain the following result:
u
Sijkl = Sijkl
dr
−
− [S gr − S dr ]ijmn [(φ(S f − S p ) + S dr − S gr )−1 ]mnos [S gr − S dr ]oskl , (2.19)
where [. . .]ijkl denotes corresponding components of tensors obtained after apply-
ing tensor operations indicated in the square brackets. The derivation of equa-
tion (2.19) has not explicitly used the fact that the material infilling the pore space
is a fluid. However, equations (2.16), (2.17) and (2.18) imply a possibility of using
f f
homogeneously distributed tensors σi j and Sijkl for describing the strain–stress rela-
tions in this material. For a pore pressure in a homogeneous fluid infilling the
pore space this is really the case. Ciz and Shapiro (2007) heuristically assumed
that such uniformly distributed tensors can, on average, describe the strain–stress
relations even in the case of a solid infilling material. Under such an assumption,
equation (2.19) can be used to approximately compute compliance tensors in such
situations, where the pore space is filled by a solid or nearly solid material like a
cement, a gas hydrate or an oil with an extremely high viscosity (e.g. a heavy oil).
In Section 2.2.2 we will pay more attention to undrained systems.
2.2 Linear stress–strain relations in fluid-saturated materials

In the rest of this book we will concentrate on the classical case, where the infill
material of the pore space is a fluid. Therefore, from this point on we must set
f
σi j = −Pp δi j . Then, Hooke’s laws (2.14) and (2.15) for the strains of the sample
and of the pore space take the following forms, respectively:
gr
i j = Sijkl
dr
σkl + (Sijkk
dr
− Sijkk )Pp , (2.20)
1 1 dr gr 1 dr gr p
− ζi j = (Sklij − Sklij )σkl + (Sijkk − Sijkk − φ Sijkk )Pp . (2.21)
Vp φ φ
2.2.1 Independent confining stress and pore pressure

Quantity −ζi j /V p expresses the strain of the pore space as an independent body.
However, the pore space along with the pore fluid, which is assumed to saturate
the pore space completely, is a part of the porous sample. The following quantity
is more natural for describing deformations of the pore space as a part of the rock
sample:
ζi j
ξi j = − . (2.22)
V
We further consider the quantity ξi j under conditions of a pore pressure acting
on the pore-space surface and a confining stress acting on the outer surface of the
sample. We also assume a homogeneous skeleton material. This last assumption
(implying that equation (2.11) is valid) is often adequate. Elastic moduli of com-
mon minerals are of the same order of magnitude and at least an order of magnitude
higher than the bulk moduli of common pore fluids. Thus, to a first approximation,
spatial variations of grain bulk and shear moduli can be neglected. This allows us
to introduce in this section several important relations. Hooke’s law for the quantity
ξi j is obtained from equation (2.21):
gr gr gr
ξi j = (Sklij
dr
− Sklij )σkl + (Sijkk
dr
− Sijkk − φ Sijkk )Pp . (2.23)
On the other hand the following dilatational strain of the fluid f results from a
homogeneously distributed pressure in it:
f = −C f Pp , (2.24)
where C f is the bulk compressibility of the fluid. In respect to the sample volume
this is the following strain (note the difference in the strain notations f and ε f ):
ε f = −φC f Pp . (2.25)
Therefore, the dilatational characteristic of the pore space
χ = ξii − ε f = ξii + φC f Pp (2.26)
expresses volumetric deformations of the pore space due to additional fluid-mass

migration only (because the effect of the fluid dilatation has been eliminated by the
second terms of the right-hand sides of the equation above).
A demonstration of this fact follows from equations (7.50)–(7.53) of Jaeger
et al. (2007). Let us consider a change of the fluid volume in a small sample of
a porous fluid-saturated material. We denote a variable mass of the fluid in the
sample by m f . The fluid has a pressure-dependent mass density ρ f . Generally, an
incremental change of the fluid volume is produced by applying pore pressure Pp
and confining stress σkl . It is defined by a change in the fluid mass δm f and a
change in the fluid density δρ f . The change of the fluid volume coincides with the
change of the pore-space volume δV p (we consider fully saturated materials). Then
(see also equation (1.23))
mf δm f m f δρ f δm f
δV p = δ = − = − V p C f Pp . (2.27)
ρf ρf ρf2 ρf
Dividing this equation by the volume of the rock sample and taking into
account (2.22) we obtain:
δm f
ξii = − φC f Pp . (2.28)
Vρ f
Combining this with (2.26) we see that
δm f
χ= . (2.29)
Vρ f
Thus, χ is indeed equal to the relative change of the pore-space volume due to
influent or effluent fluid mass only.
The corresponding Hooke’s law is obtained from equations (2.23) and (2.26):
gr
χ = (Sklii
dr
− Sklii )σkl + (C dr − C gr − φC gr + φC f )Pp , (2.30)
where C dr and C gr are bulk compressibilities of the drained skeleton of the rock
and of its grain material, respectively.
By inverting for the stress tensors, Hooke’s law (2.20) can also be written in the
stress–strain form:
σi j = Cijkl
dr
kl − αi j Pp . (2.31)
dr
Here Cijkl ≡ [(S dr )−1 ]ijkl denotes the components of the stiffness tensor of the
drained rock skeleton. We have also introduced a new tensor with components αi j :
gr
αi j ≡ Cijmn
dr
(Smnkk
dr
− Smnkk ). (2.32)
Note that the sample strain kl can be expressed from equation (2.31) over the
confining stress and the pore pressure:
kl = Sklij
dr
(σi j + αi j Pp ). (2.33)
Thus, quantity αi j plays the role of a tensorial effective stress coefficient for the
strain of the sample (see our discussion of the effective stress in Section 2.9.6).
In an isotropic medium for tensors of compliances and stiffnesses the following
relations hold: Smnkk = δmn /(3λ + 2μ) and Cijmm = δi j (3λ + 2μ). Therefore, tensor
αi j simplifies further:
αi j = αδi j , (2.34)
where
C gr K dr
α ≡1− dr
=1− (2.35)
C K gr
is called the Biot–Willis coefficient. Here K dr and K gr are the bulk moduli of
the drained skeleton of the rock and of its grain material, respectively. Note that
definition (2.35) of the scalar quantity α is meaningful for anisotropic media too.
In isotropic media, equation (2.31) takes the following form:
σi j = λdr δi j + 2μdr i j − α Pp δi j . (2.36)
A double-dot scalar product with δi j then gives

1 1
= σii + α Pp . (2.37)
K dr 3
The quantity σii /3 is an average confining normal stress. Taking into account the
form of equation (2.37) the combination α Pp + σii /3 is sometimes called an effec-
tive stress for the sample’s dilatation ≡ ii (note the difference between the
effective stress defined previously in equation (2.6) and the effective stress for
the sample’s dilatation; we will again discuss this subject in Section 2.9.6).
Let us further simplify the dilatational quantity χ using the notation intro-
duced above. For this we substitute the confining stress from equation (2.31) into
equation (2.30):
gr
χ = (Sklii
dr
− Sklii )(Cklmn
dr
mn − αkl Pp ) + (C dr − C gr − φC gr + φC f )Pp
= αmn mn + Ma−1 Pp , (2.38)
where in the last part of this equation we have introduced a new quantity Ma :
Ma ≡ [C dr α 2 + (α − φ)C gr + φC f − αkl αmn Sklmn

dr
]−1 . (2.39)
In an isotropic medium Ma = M, where the quantity

φ α − φ −1
M≡ + (2.40)
Kf K gr
is frequently called the Biot modulus (Detournay and Cheng, 1993). From equation
(2.38) we finally obtain:
Pp = Ma (χ − αmn mn ). (2.41)
For isotropic systems this equation simplifies to the following important relation
between the pore pressure and the dilatations:
Pp = M(χ − α). (2.42)
This equation and equation (2.37) show that, in elastically isotropic media, of
the four dilatational- and pressure-type quantities , χ , Pp and σii , only two are
independent. Thus, two more relations between any three of these quantities can
additionally be written down:

1 α2 α α 1 1
χ= + Pp + σii ≡ Pp + σii , (2.43)
M K dr 3K dr K dr B 3
and
1
σii = (K dr + α 2 M) − α Mχ ≡ K u − α Mχ . (2.44)
3
In the two equations above we have introduced the bulk modulus of an undrained
rock K u and a coefficient B (called also Skempton’s coefficient). We will discuss
these quantities in the next section. They are explicitly given by equations (2.51)
and (2.54), respectively.
For some rock samples, drained and grain bulk moduli as well as Poisson’s
ratios and other poroelastic constants are given in table 7.2 of the book by Jaeger
et al. (2007).
2.2.2 Linear stress–strain relations in undrained media

Let us again return to a quite general situation of a possibly heterogeneous
anisotropic grain material, and consider a special case of undrained systems
(see Figure 2.5). Later we will see that seismic body waves can usually be
considered as propagating in undrained rocks.
In the case of an undrained system, when the fluid has no possibility of migrating
(at least on average) relative to the solid frame (i.e. the mass of the material infilling
Figure 2.5 A jacketed porous sample representing an undrained system.

the pore space is constant), the strains ε f and −ζii /V p must be equal (i.e. the
quantity χ is vanishing in equation (2.26)), and thus equations (2.21) and (2.25)
yield:
gr
− φC f Pp = (Sklii
dr
− Sklii )σkl + (C dr − C gr − φC p )Pp , (2.45)
where we recall that C dr , C gr , C p are bulk compressibilities of the drained skeleton

of the rock, of its grain material and of its pore space. Now the pore pressure can
be calculated:
1
Pp = − Bkl σkl , (2.46)
3
where we have introduced a tensor with components Bkl :
gr
dr
Sklii − Sklii
Bkl ≡ 3 . (2.47)
C dr − C + φ(C f − C p )
gr
This tensor relates pore-pressure changes in an undrained anisotropic sample to a

confining stress tensor causing these changes.
u
The undrained compliance of the saturated porous rock Sijkl can also be calcu-
lated. For this strain (2.20) must be rewritten by explicitly using the undrained
compliance and the confining stress:
gr
u
Sijkl σkl = Sijkl
dr
σkl + (Sijkk
dr
− Sijkk )Pp . (2.48)
Substituting pore pressure Pp from equation (2.46) into equation (2.48) and
excluding tensor σkl we obtain the following result:
gr gr
(Smmij
dr
− Smmij )(Snnkl
dr
− Snnkl )
u
Sijkl = Sijkl
dr
−
φ(C f − C p ) + C dr − C gr
1 gr
= Sijkl
dr
− Bi j (Snnkl
dr
− Snnkl ). (2.49)
3
This equation was first derived by Brown and Korringa (1975).
Let us assume an orthorhombic symmetry (including isotropy and transverse
isotropy) of the drained skeleton and grain materials with coinciding symmetry
planes. Then quantities of the type Smmij will vanish if i and j do not coincide.
Therefore, the following components of the tensors of compliances and stiffnesses
of undrained rocks (in contracted notations) s44 , s55 , s66 , c44 , c55 , c66 are not affected
by the presence of fluids in pores. These components coincide with corresponding
quantities of the drained skeleton. They describe dynamics of shear deformations
of the sample, and thus logically are independent of the pore fluid. Note that owing
to the same reasons in orthorhombic media, matrix Bkl , defined by equation (2.47),
is also diagonal.
If all materials are homogeneous and isotropic, then a twofold double-dot prod-
uct of the Brown–Korringa equation (2.49) with δi j and δkl will give the famous
Gassmann equation (Gassmann, 1951):
(C dr − C gr )2
C u = C dr − . (2.50)
φ(C f − C gr ) + C dr − C gr
This equation is the basis of fluid-substitution calculations broadly applied in reser-
voir geophysics. Note that here we set C p = C gr . The Gassmann equation is
frequently written in terms of the corresponding bulk moduli, i.e. the bulk mod-
ulus of the undrained (i.e. fluid-saturated) medium K u , the bulk modulus of the
drained medium (i.e. of the skeleton) K dr , the bulk modulus of the grain material
(i.e. of a material of which the skeleton “is made,” frequently also called a solid
material) K gr , and the bulk modulus of the fluid K f :
K u = K dr + α 2 M, (2.51)
where M is the Biot modulus defined by equation (2.40). For any type of saturating
fluid the bulk modulus of a saturated rock K u must be larger than the bulk modulus
of the rock skeleton K dr . Thus, the modulus M must be always positive. Therefore,
the following restriction must be always valid: α ≥ φ (see equation (2.40)). An
even more-strict limitation for α can be obtained from the upper bound of the
quantity K dr if the rock skeleton is considered as a composite material of porosity
φ (see Zimmerman et al., 1986; see also our discussion later in Section 2.9.4).
In an undrained homogeneous porous sample an applied confining pressure Pc
causes changes of the pore pressure. The general equation (2.46) will describe this
if we substitute σkl = −Pc δkl and make corresponding simplifications
1
Pp = B Pc = − Bσkk , (2.52)
3
where the quantity
C dr − C gr 1
B≡ = Bii (2.53)
C dr − C + φ(C − C )
gr f gr 3
is called the Skempton coefficient. In isotropic rocks the Skempton coefficient can
be further simplified:
αM
B= . (2.54)
Ku
Note that we have already used this relation in equation (2.43).
Finally, in undrained systems χ = 0, and equation (2.41) yields
Pp = −Ma αmn mn . (2.55)

In isotropic systems this equation simplifies to

Pp = −Mα. (2.56)
Substituting relation (2.55) into equation (2.31) yields
σi j = Cijkl
u
kl , (2.57)
u
where the stiffness tensor of an undrained rock Cijkl is given by
u
Cijkl = Cijkl
dr
+ αi j αkl Ma . (2.58)
In undrained isotropic systems the substitution of (2.56) into (2.36) yields the
following form of Hooke’s law:
σi j = (λdr + α 2 M)δi j + 2μdr i j , (2.59)
which is consistent with the Gassmann equation (2.51) and with a well-known
result (already discussed above in a general form for orthorhombic media) for the
undrained shear modulus:
μu = μdr . (2.60)
2.3 Fluid flow and dynamic poroelasticity

Our consideration above was restricted to static poroelastic phenomena. Dynamic
effects are given by a time-dependent balance of forces (in other words, momen-
tum conservation or Newton’s second law) acting on a representative volume of
a porous fluid-saturated rock. Therefore, a description of the rock dynamics must
also take into account processes related to fluid flows.
We continue to consider a differentially small (but representative) elementary
volume of a fluid-saturated rock. We introduce a velocity vector q of an average
displacement of the volume of the pore fluid with respect to the skeleton (this
displacement is sometimes called a global flow). This vector is also called the fluid
flux vector (see Jaeger et al., 2007) or the vector of filtration velocity. Its component
qi is equal to a volume of the fluid crossing in a unit time a unit planar surface of
the rock (attached to the solid skeleton) with a normal directed along axis x̂i . Let
us further denote a vector of the corresponding average displacement of the fluid
volume (relative to the solid skeleton) as w = (w1 , w2 , w3 ). Therefore,
∂w
= q. (2.61)
∂t
We recall that q describes a motion of the volume of fluid in the pore space.
Therefore, according to definition (2.61) we can also write
w = φ(uf − u), (2.62)
2.3 Fluid flow and dynamic poroelasticity 63
where uf is a displacement vector of an average (representative) fluid particle rel-

ative to its initial location (in a coordinate system common for the solid as well
as the fluid). We recall also that u is a displacement of an average (representative)
particle of the solid relative to its initial location. It follows from (2.61) that the
dilatation ∂wi /∂ xi is equal to a fluid-volume decrement in the corresponding ele-
mentary volume of the rock. Corresponding to the definition (2.26) of the quantity
χ we must have:
∂wi
= −χ . (2.63)
∂ xi
Taking the time derivative at the left- and right-hand sides of this equation, we
obtain
∂χ ∂qi
+ = 0. (2.64)
∂t ∂ xi
This equation expresses the fluid-mass conservation in terms of migration of an
additional fluid volume per unit time necessary to compensate temporal changes of
the pore-space volume per unit volume of the rock. This is a form of the so-called
fluid-continuity equation (Jaeger et al., 2007).
Taking into account the fluid-mass component permanently remaining in the
pore space during a deformation process, the continuity equation can be rewritten
in terms of masses explicitly (Coussy, 2004; Segall, 2010):
∂ρ f φ ∂ρ f qi
+ = 0, (2.65)
∂t ∂ xi
where the first term on the left-hand side describes temporal changes of a fluid
mass per unit volume of the rock, and the second term describes divergence of the
vector of the fluid-mass flux.
An experimental observation called Darcy’s law relates a stationary filtration
velocity of a laminar fluid flow in a porous system to a gradient of the pore pressure:
ki j ∂ Pp
qi = − , (2.66)
η ∂x j
where η is the dynamic viscosity of the pore fluid. This equation defines a tensor
with components ki j , which is called the tensor of permeability. Here (and gen-
erally, if not otherwise specified) the gravity force is neglected. Darcy’s law can
be also written in a form explicitly expressing an equilibrium between elastic and
friction forces acting on the fluid:
∂ Pp
η[k −1 ]i j q j = − , (2.67)
∂ xi
where [k −1 ]i j denotes corresponding components of the tensor inverse to the tensor

of permeability.
Summarizing, equation (2.31) describes the elastic stress acting on the rock.
Its spatial derivatives represent elastic forces. Equation (2.41) describes the fluid
pressure. Its gradient gives a fluid driving force. The friction force acting on the
fluid participates in Darcy’s law (2.67). Biot (1962) combined these equations into
the following system of equations describing the balance of dynamic forces applied
to an elementary representative volume of a fluid-saturated rock:
∂ dr ∂2 ∂
Cijkl + Ma αi j αkl kl − αi j Ma χ = ρ 2 u i + ρ f qi , (2.68)
∂x j ∂t ∂t
∂ ∂2 ∂
− (Ma (χ − αmn mn )) − η[k −1 ]i j q j = ρ f 2 u i + ρ f m i j q j . (2.69)
∂ xi ∂t ∂t
Here the first equation expresses a force balance for a small (but still representa-
tive) element of a fluid-saturated rock as a whole. The second equation expresses a
force balance for the fluid in this rock element. The left-hand sides of these equa-
tions express the elastic and friction forces. Their right-hand sides describe inertial
forces. We have introduced here some new notations: ρ is the total density of the
fluid-saturated rock; the matrix m i j (here made dimensionless) was introduced by
Biot (1962) as a volume-averaged matrix product aki ak j , where aki (r) is a matrix
coefficient relating a vector field of the local micro-velocity of the fluid in pores (in
respect to the skeleton) to the global (macro) filtration-velocity vector qi .
System (2.68)–(2.69) is a system of six equations for all displacement and
filtration-velocity components. It is restricted in its validity for a not too high fre-
quency range, where the fluid flow in pores can be assumed to be of Poiseuille type
(laminar viscous flow in a tube). Biot (1956) estimates that one quarter of the shear-
wave length (the boundary layer) in a viscous fluid should not become shorter than
the pore diameter d p . This leads to the following restriction of the frequency range:
π 2η
ω < ωl ≡ , (2.70)
2d 2p ρ f
where ω is an angular frequency of a wave motion and ωl is a characteristic

frequency of the Poiseuille flow. For example, assumming rather large pores of
d p = 10−6 m and a saturating flud with physical properties close to those of the
water, ρ f = 103 kg/m3 and η = 10−3 Pa·s, we obtain ωl ≈ 5 · 106 . Therefore, the
characteristic frequency of the Poiseuille flow is very high, of the order of MHz.
It becomes even higher for smaller pores. If condition (2.70) is not satisfied, the
viscosity, the permeability and the matrix m i j will become frequency dependent.
However, this fact is not significant for our further consideration, which will be
restricted later to a much lower frequency range.
Equation system (2.68)–(2.69) describes four wavefields independently prop-
agating in a homogeneous anisotropic poroelastic continuum (Carcione, 2007).
Three of them are analogous to the three elastic waves in a general anisotropic
solid. The fourth one corresponds to Biot’s slow wave, which we will discuss later
in detail.
To discuss the wave motions we first assume elastic and hydraulic isotropy and
homogeneity of rocks. The hydraulic isotropy is given by statistical isotropy of
the pore space. This leads to statistical isotropy of the fluid micro-velocity field.
Therefore, the following relation holds: ki j = kδi j , where k is a scalar permeability
of the rock. Further, m i j = T δi j /φ, where T ≥ 1 is a parameter (called tortuosity)
of the order of a mean square of a ratio of two lengths: a trajectory length of a
pore-fluid particle between two points of the pore space to a straight line distance
between these two points (see Carcione, 2007, and references therein). Equations
(2.68)–(2.69) then simplify to the following ones:
∂ ∂χ ∂i j ∂2 ∂
(λdr + Mα 2 ) − αM + 2μdr = ρ 2 u i + ρ f qi , (2.71)
∂ xi ∂ xi ∂x j ∂t ∂t
∂ ∂2 T ∂ η
− (M(χ − α)) = ρ f 2 u i + ρ f qi + qi . (2.72)
∂ xi ∂t φ ∂t k
This system of equations describes propagation of two dilatational waves and one
shear (rotational) wave (see Biot, 1956). These equations explicitly contain a dilata-
tion of a porous fluid-saturated elementary rock volume = ∇u (equal to the
dilatation of the skeleton) and a dilatation of the pore space of this elementary rock
volume caused by a fluid flow in or out of this volume: χ = −∇w.
2.4 Poroelastic wavefields

To analyze possible wave motions in more detail we apply the divergence operator
to equations (2.71) and (2.72). Taking into account that
∂ 2 i j ∂ 2
= 2 = ∇ 2 , (2.73)
∂ xi ∂ x j ∂ xi
and using also continuity equation (2.64), we obtain
∂2 ∂2
∇ 2 [H − α Mχ ] = ρ − ρ f χ, (2.74)
∂t 2 ∂t 2
∂2 T ∂2 η ∂
∇ 2 (Mα − Mχ ) = ρ f − ρ f χ − χ. (2.75)
∂t 2 φ ∂t 2 k ∂t
We have introduced here a poroelastic modulus
H ≡ λdr + 2μdr + Mα 2 . (2.76)
This quantity is equal to the P-wave modulus of isotropic undrained rocks. This
can be seen from a comparison of equation (2.76) with equations (2.51) and (2.60).
Let us also consider rotational (shear) wave motions. For this we apply the curl
operator (also called the rotor operator) to equations (2.71) and (2.72). This oper-
ator is equal to a vectorial product of the vector ∇ with the vectors on the left-
and right-hand sides of these equations. Taking into account that ∇ × ∇ = 0,
∇ × ∇χ = 0 and that 2∂i j /∂ x j is the ith component of the vector ∇ 2 u + ∇(∇u)
(see equations (1.43)–(1.46)), we obtain:
∂2 ∂
μdr ∇ 2 ∇ × u = ρ ∇ × u + ρ f ∇ × q, (2.77)
∂t 2 ∂t
∂2 T ∂ η
0 = ρf ∇ × u + ρf ∇ × q + ∇ × q. (2.78)
∂t 2 φ ∂t k
Comparing both equation systems, (2.74)–(2.75) and (2.77)–(2.78), we observe
that they are mutually independent. Therefore, these two equation systems
describe independent wavefields, respectively. Indeed, dilatational quantities are
not involved in equations (2.77)–(2.78). Equally, rotational quantities are not
involved in equations (2.74)–(2.75). Thus, dilatational and rotational motions
propagate mutually independently.
To further analyze wave motions it is convenient to write equation sys-
tems (2.74)–(2.75) and (2.77)–(2.78) in terms of the displacement vector fields
u and w. The vector fields u and w can always be represented as sums of their
dilatational and rotational vector-field parts ud , ur and wd , wr , respectively. Thus:
u = ud + ur ,
w = wd + wr ,
∇ × ud = 0,
(2.79)
∇ur = 0,
∇ × wd = 0,
∇wr = 0.
Taking into account these decompositions and the fact that, if both the vector
product and the scalar product of the vector ∇ with a vector field are equal to zero
everywhere in space, then such a vector field (which must be also equal to zero at
infinite distances from its source) will be equal to zero too, we obtain two other
equation systems, respectively:
∂2 ∂2
∇ 2 [H ud + α Mwd ] = ρ u d + ρ f wd (2.80)
∂t 2 ∂t 2
∂2 T ∂2 η ∂
∇ 2 (Mαud + Mwd ) = ρ f u d + ρ f wd + wd (2.81)
∂t 2 φ ∂t 2 k ∂t
and
∂2 ∂2
μdr ∇ 2 ur = ρ u r + ρ f wr (2.82)
∂t 2 ∂t 2
∂2 T ∂2 η ∂
0 = ρf u r + ρ f wr + wr . (2.83)
∂t 2 φ ∂t 2 k ∂t
We observe that in an infinite homogeneous isotropic porous fluid-saturated
medium, dilatational and rotational parts of the displacement fields are described
by completely independent equation systems. Thus, perturbations of these field
parts propagate mutually independently. Because equation systems (2.80)–(2.81)
and (2.82)–(2.83) describe the displacement fields completely, then they describe
all possible wave motions. Therefore, an infinite homogeneous medium can inde-
pendently propagate dilatational and rotational waves only. Note that additional
boundary conditions (for example, a free surface) can lead to the emergence of
additional independent wave motions (for example, surface waves), where the
dilatational and rotational displacement fields can be coupled. Such wavefields also
are solutions of the initial equation system (2.71) and (2.72). They are beyond the
scope of this book.
2.4.1 Dispersion relations for poroelastic wavefields

Any wavefield can be decomposed (by Fourier transformation) into a superposition
of time-harmonic plane waves. Thus, further analysis of possible wave motions is
performed by substituting into equation systems (2.80)–(2.81) and (2.82)–(2.83)
trial wavefield solutions in the forms:
ud = ud0 ei(ωt−κ d r) ,
wd = wd0 ei(ωt−κ d r) ,
(2.84)
ur = ur0 ei(ωt−κ r r) ,
wr = wr0 ei(ωt−κ r r) .
Here κ d and κ r are unknown wave vectors and ud0 , wd0 , ur0 and wr0 are unknown
polarization vectors of the corresponding plane waves.
We restrict our consideration of wave motions to isotropic systems. Thus, all
propagation directions are equivalent. For example, we can consider plane waves
propagating along the x-direction (i.e. x̂1 -direction). Then the scalar products κ d r
and κ r r can be replaced by κd x and κr x, respectively. Here κd and κr are corre-

sponding wave numbers (i.e. x-projections of the wave vectors). Further, because
of the conditions (2.79), the polarization vectors of the dilatational waves must
be parallel to x̂1 and can be replaced by the corresponding vector components,
u d0 and wd0 . The rotational wave motions must be polarized orthogonally to the
propagation direction. Moreover, equation (2.83) indicates that the vectors ur0
and wr0 must be parallel. We can choose the y-direction (i.e. x̂2 -direction) to be
parallel to the polarization of the rotational waves and replace the rotational polar-
ization vectors by their corresponding vector components u r 0 and wr 0 . Taking these
simplifications into account we obtain:
0 = (κd2 H − ω2 ρ)u d0 + (κd2 α M − ω2 ρ f )wd0 (2.85)
T η
0 = (κd2 Mα − ω2 ρ f )u d0 + (κd2 M − ω2 ρ f + iω )wd0 (2.86)
φ k
and
0 = (κr2 μdr − ω2 ρ)u r 0 − ω2 ρ f wr 0 (2.87)
T η
0 = ρ f u r 0 + (ρ f − i )wr 0 . (2.88)
φ kω
These are two linear algebraic equation systems for wavefield polarizations u d0 ,
wd0 , u r 0 and wr 0 . They will have non-zero solutions only if the corresponding
determinants are equal to zero, respectively:
T η
0 = (κd2 H − ω2 ρ)(κd2 M − ω2 ρ f + iω ) − (κd2 α M − ω2 ρ f )2 (2.89)
φ k
and
T η
0 = (κr2 μdr − ω2 ρ)(ρ f − i ) + ω2 ρ 2f . (2.90)
φ kω
These are dispersion equations relating wave numbers and frequencies of possible
wave motions (note the analogy to equation (1.33)).
We consider first rotational wave motions. Solving equation (2.90) we obtain the
following wave number:
ρ ρfφ φη −1
κr2 = ω2 (1 − (1 − i ) ). (2.91)
μdr Tρ ρ f T kω
For frequencies significantly below the critical frequency (also called the
characteristic frequency of the global flow)
φη
ωc ≡ (2.92)
ρf Tk
the absolute value of the quantity (1 − iηφ/(ρ f T kω))−1 is much smaller than 1,
and we will obtain κr2 = ω2 ρ/μdr . This gives the following phase velocity:
μdr
cs = . (2.93)
ρ
This is a velocity of a seismic shear wave in an undrained isotropic rock (note the
difference in the density of an undrained medium and of its drained skeleton).
Usually for realistic rocks and fluids the critical frequency ωc /(2π) is of the
order of 105 Hz or higher. It can be even higher than the frequency limit of a
laminar flow in pores (see equation (2.70)). Therefore, in such fields as borehole
seismology, reflection seismology, earthquake seismology and hydrogeology, the
system of poroelastic equations can be considered in the low-frequency (in respect
to ωc ) range.
The low-frequency range has the following meaning in terms of forces acting
on the fluid in the pore space. In the low-frequency range viscous forces dominate
over inertial forces (acting on the fluid due to wave-motion-related variations of
velocities of the solid skeleton). In the high-frequency range the situation is oppo-
site. Accelerations of the solid skeleton become so high that they dominate over
the viscous friction between the solid and the fluid.
In the low-frequency range, all terms of order ω2 can be neglected with respect
to terms of order ω and lower. Correspondingly, equation (2.89) simplifies to
η
0 = (κd2 H − ω2 ρ)(κd2 M + iω ) − (κd2 α M − ω2 ρ f )2 . (2.94)
k
This is a bi-quadratic equation having two solutions for κd2 . Thus, we observe a
possibility of two types of dilatational waves. The wave number can be of the order
of ω or of a lower ω-order. Higher orders in ω are excluded in the low-frequency
range. We first assume that κd = O(ω). Then the second term in the equation
above is of the order O(ω4 ). It can be neglected in respect to the first one (which
is O(ω3 )), and we obtain
0 = κd2 H − ω2 ρ. (2.95)
Thus, the phase velocity of this first dilatational wave is:
H
cP I = . (2.96)
ρ
This is a velocity of a seismic longitudinal wave in an undrained isotropic rock

(compare this expression to equations (2.76), (2.51) and (2.60)). Thus, the quantity
H is indeed the modulus of seismic P-waves.
Further, we consider a possibility that the wave number can be of a lower order
than O(ω). Then we can neglect terms O(ω2 ) in respect to the terms O(ω) and
O(κd2 ), and equation (2.94) simplifies to:
η
0 = κd2 H (κd2 M + iω ) − (κd2 α M)2 . (2.97)
k
Solving this equation we obtain:
1
κd2 = −iω , (2.98)
D
where
k
D= (2.99)
ηS
and
H
S= . (2.100)
(λdr + 2μdr )M
Quantity S is called the storage coefficient (see Jaeger et al., 2007, p. 188) or
sometimes also the uniaxial storage coefficient. The coefficient D is called the
hydraulic diffusivity.
For plane waves propagating in the direction of positive x-values, equa-
tion (2.98) provides us with the following wave number:
ω
κd = (1 − i) . (2.101)
2D
We observe that the absolute values of the real and imaginary parts of the
wave number are equal. Such a feature is typical for so-called diffusion waves
(Mandelis, 2000) which are characterized by a very strong dissipation of their
(initially elastic) energy. We will consider this in more detail. Frequency divided
by the real part of the wave number yields the phase velocity of the corresponding
wave. The imaginary part of the wave number is equal to the attenuation coeffi-
cient of the wave. Indeed, the corresponding dilatational plane wave (2.84) has the
following form:
√ √
2Dω) −x ω/(2D)
ud = ud0 eiω(t−x/ e ,
√ √ (2.102)
2Dω) −x ω/(2D)
wd = wd0 eiω(t−x/ e .
Therefore, for the phase velocity c P I I of this second dilatational wave we obtain
√
c P I I = 2Dω = 4π D/t, (2.103)
where the last part of this equation yields the phase velocity of the second dilata-
tional wave with a time period t (i.e. with an angular frequency ω = 2π/t). The
wavelength of the second dilatational wave λ P I I is given by
2π 8π 2 D √
λP I I = cP I I = = 4π Dt. (2.104)
ω ω
Here again the last part of the equation yields the wavelength of the second
dilatational wave with time period t (angular frequency ω = 2π/t).
The attenuation coefficient α P I I of the second dilatational wave (see the second
exponential factor on the right-hand side of equations (2.102)) is given by
ω π
αP I I = = . (2.105)
2D Dt
The last part of equation (2.105) yields the attenuation coefficient of the second
dilatational wave with time period t and the angular frequency 2π/t, respectively.
Therefore, over a distance equal to a single wavelength the amplitude of this
wave decreases in exp 2π ≈ 535 times. Over a distance of a single wavelength the
energy of the second dilatation wave will be nearly completely transformed into
heat. Simultaneously we observe that the wave numbers of the rotational and first
dilatational waves do not have imaginary parts. Therefore, for frequencies much
smaller than ωc the attenuation of these waves is vanishingly small.
2.4.2 Particle motions in poroelastic wavefields

To clarify the reason for such contrasting behavior of the wave motions we will
consider their polarizations. First we consider again the system of equations (2.87)–
(2.88). Recalling in the first one that κr2 = ω2 ρ/μdr , and in the second one that we
consider the low-frequency limit, we obtain wr 0 = 0. In other words, in rotational
wave motions, particles of the solid and of the fluid have the same displacements.
They move together and do not move with respect to each other. Analogous consid-
eration of equation system (2.85)–(2.86) along with equation (2.95) yields for the
first dilatational wave wd0 = 0. Note that also in equation (2.86) we must consider
the low-frequency limit, i.e. the last term with factor w0 is dominant and it must
vanish in this limit. Thus, also in this wave, both fluid and solid particles move with
equal displacements and do not move in respect to each other. Such a polarization
feature of these two wave types is responsible for their vanishing attenuation in the
low-frequency range.
We consider again equation system (2.85)–(2.86). For the second dilatational
wave, where the wave number satisfies relation (2.98), both equations provide the
same result (again, higher-order terms in ω must be neglected):
H u d0 = −α Mwd0 . (2.106)
Substituting here equation (2.62) we obtain:

f
(H − φα M)u d0 = −αφ Mu d0 . (2.107)
Both quantities, H − φα M = λdr + 2μdr + α(α − φ)M and αφ M, are always

positive (because α ≥ φ; see also our discussion after equation (2.51)). Thus,
in this case, wave particles of the fluid and of the solid move in opposite direc-
tions. This produces a strong friction loss and leads to a strong attenuation of
the wave.
Summarizing, in a homogeneous isotropic fluid-saturated poroelastic continuum
there are three types of waves propagating a perturbation from a source to a point
of observation. These are two types of elastic body waves, P and S (these are usual
longitudinal and shear seismic waves), and a highly dissipative second dilatational
(longitudinal) wave. Usually the phase velocity of the latter is very small. It tends
to zero in the zero-frequency limit. Below the critical frequency the second dilata-
tional wave has the smallest phase velocity from the three types of waves. Thus,
frequently this wave is also called the slow wave.
2.4.3 Fluid flow and attenuation of seismic waves

Fluid flows can strongly contribute to the attenuation of seismic waves. In the the-
oretical model of a homogeneous poroelastic medium (as described earlier in this
section) the dissipation of seismic waves (a very insignificant one) is caused by
the global flow, i.e. a relative movement between the solid matrix and the fluid on
the scale of a wavelength. In a heterogeneous medium, additional flow phenomena
may cause additional (sometimes very significant) dissipation of the elastic energy.
For example, seismic wavefields in a stack of layers with variable compliances
cause a flow of the pore fluid (also called mesoscopic flow, local flow or inter-
layer flow) across interfaces from more compliant layers into stiffer layers and vice
versa (White, 1983; Norris, 1993; Gurevich and Lopatnikov, 1995; Gelinsky and
Shapiro, 1997; Pride et al., 2004). Indeed, the fluid pressure tends to equilibrate
between adjacent heterogeneities (e.g., layers) by a motion of the viscous pore
fluid across the boundaries of heterogeneities. This fluid flow produces strong fric-
tional losses, and it is a reason for the strong dissipation of seismic waves. The
phenomenon of the mesoscopic fluid flow can be also described as an excitation
of diffusion slow waves at interfaces of heterogeneities by scattering of the normal
seismic P- (and also sometimes S-) waves.
The mesoscopic flow is especially significant at the characteristic frequencies
where the scale of rock heterogeneities is comparable with the wavelength of the
slow wave. Such a flow is possibly one of main causes of the attenuation and
dispersion of seismic waves in the reflection seismic frequency range, 10–100 Hz
(Shapiro and Hubral, 1999). The various attenuation mechanisms are approxi-
mately additive, dominated by the mesoscopic flow at low frequencies and in the
seismic frequency range. Elastic scattering is significant over a broad frequency
range from seismic to sonic frequencies. The global flow contributes mainly in the
range of ultrasonic frequencies. One more type of fluid flow, the so-called squirt
flow, seems to be mainly responsible for seismic-wave attenuation in the frequency
range of several 1000 Hz. It takes place due to fluctuation of compliances of the
pore space caused by its micro-scale geometry (see Gurevich et al., 2010, and liter-
ature cited therein). The squirt can be also understood as a fluid flow between stiff
and compliant parts of the pore space (see our discussion at the end of this chapter).
From the seismic frequency range up to ultrasonic frequencies, attenuation of seis-
mic waves due to heterogeneity is strongly enhanced compared with that in a
homogeneous poroelastic medium.
2.4.4 Slow wavefields in the low-frequency range

Practical experience shows that a cloud of fluid-induced microseismic events
requires hours or even days to reach a size of several hundred meters. This is
definitely too slow a process to be attributed to elastic-wave propagation. Elas-
tic waves in well consolidated rocks propagate during seconds on kilometer-scale
distances. However, elastic waves are primarily responsible for the elastic stress
equilibration. This indicates that the triggering of at least some microseismic events
has to be related to the slow wave.
In the following discussion we will derive a wave equation for slow waves. Equa-
tion (2.98) shows that, in the range of frequencies below the critical one, the wave
number of slow waves is of order O(ω1/2 ). This means that for slow wavefields
time derivatives can be neglected in respect to the same- or lower-order spatial
derivatives. Equivalently, inertial terms can be neglected in respect to static and
dissipative terms because inertial forces are negligible in comparison to the viscous
friction and elastic forces.
We simplify equations (2.71) and (2.72) according to this approximation
(assuming initially an elastic and hydraulic isotropy and homogeneity of rocks):
∂ ∂χ ∂i j
(λdr + Mα 2 ) − αM + 2μdr = 0, (2.108)
∂ xi ∂ xi ∂x j
∂ η ∂
− (M(χ − α)) = wi . (2.109)
∂ xi k ∂t
We recall that the slow wavefields are dilatational ones. In other words, the dis-
placement fields u = ud and w = wd are irrotational (see equations (2.79)). Taking
also into account that the third term on the left-hand side of equation (2.108) can
be rewritten by using relations (1.43), (1.48) and ∇ 2 ud = ∇ we can simplify this
equation further:
∂ ∂χ
H − αM = 0. (2.110)
∂ xi ∂ xi
Taking this into account by applying the divergence operator to equation (2.109)
we obtain a closed equation describing slow wavefields:
∂
D∇ 2 χ = χ. (2.111)
∂t
This is a diffusion equation for the volume of filtrating fluid per unit volume of
rocks. It is not surprising that the “wave” equation for the slow wavefields is a
diffusion one. Above we have already seen that the slow waves are waves of a
diffusion type.
If we consider slow wavefields in infinite media, where at infinity both perturba-
tions, and χ , disappear, we obtain from equation (2.110):
αM
= χ. (2.112)
H
Then, taking into account equation (2.42) we obtain that in slow wavefields
χ
Pp = . (2.113)
S
This transforms the diffusion equation for χ into the diffusion equation for the pore
pressure:
∂
D∇ 2 Pp = Pp . (2.114)
∂t
Clearly, the same equation will be valid also for . Therefore, the slow wavefields
describe well-known and simple physical phenomena. These are pore-pressure
diffusion and fluid-mass diffusion in a porous rock.
The slow wavefields can also be expressed in terms of diffusion-like relaxation
of elastic stresses. For this we must combine equation (2.44) with equation (2.112).
This yields:
4μdr α M
σii = − χ. (2.115)
H
Thus, in the slow wavefield the same diffusion equation (2.114) is also valid for the
quantity σii .
Let us now consider a more general situation: elastically homogeneous and
isotropic rocks, which can be hydraulically anisotropic and heterogeneous. Then,
for the slow wavefields equations (2.68) and (2.69) will simplify so that equa-
tion (2.108) will be still valid and equation (2.69) will transform into
ki j ∂ ∂
− (M(χ − α)) = wi . (2.116)
η ∂x j ∂t
Applying the divergence operator yields
∂ ki j ∂ ∂
(M(χ − α)) = χ . (2.117)
∂ xi η ∂ x j ∂t
It is clear that, if we look for a solution of equation system (2.108) and (2.117)
under the separation assumption (2.79), we will obtain a closed equation sys-
tem (2.110) and (2.117) describing irrotational wavefields only. Therefore, we
observe that in hydraulically heterogeneous and anisotropic media dilatational slow
wavefields propagate independently from other wave motions. Moreover, these two
equations can be combined into the following one:
∂ ∂ ∂
Di j χ = χ, (2.118)
∂ xi ∂x j ∂t
where
ki j
Di j = (2.119)
ηS
is a tensor of hydraulic diffusivity. This is a quite general equation of diffusion of
the dilatational perturbation χ .
Assuming finally that we consider slow wavefields in an infinite medium with
wavefields vanishing at infinity, we will still be able to apply equation (2.113) and
thus:
∂ ∂ ∂
Di j Pp = Pp . (2.120)
∂ xi ∂x j ∂t
We observe that the dynamic equation describing slow wavefields in infinite elas-
tically homogeneous and isotropic but hydraulically heterogeneous and anisotropic
media is the pressure-diffusion equation (2.120). The diffusion equation (2.118) is
more general. It does not require wavefields to be vanishing at limiting boundaries
of a finite domain.
2.4.5 Elastic P- and S-waves in the low-frequency range

Let us briefly consider a possibility of the wave numbers to be of order O(ω). As
we have seen above this is a feature of normal seismic waves. Then, in the limit
of low frequencies in equation (2.68) we must keep all terms, because they are all
of the order O(ω2 ). However, equation (2.69) provides great simplifications. The
leading term there (the last one on the left-hand side of the equation) is of the order
O(ω). It is proportional to the filtration velocity q. Thus, for normal seismic waves
in the low-frequency range,
q = 0. (2.121)
This is not surprising. In the low-frequency range the influence of the global fluid
flow on the propagation of normal seismic waves is vanishing. In other words, the
relative motions of the fluid and of the solid in a normal seismic wave is vanishing.
We have already seen this above by considering polarizations of poroelastic plane
waves. However, this also means that χ = 0. This leads further to the statement
that in the normal seismic wavefields the following relation is valid (see equations
(2.42), (2.43) and compare with (2.52) for an undrained sample):
Pp = −α M = −Bσkk /3. (2.122)
Equation (2.68) can be then simplified:
∂ u ∂u k ∂ 2ui
Cijkl =ρ 2 . (2.123)
∂x j ∂ xl ∂t
This is the Lamé equation with parameters of the undrained rock (compare to equa-
tion (1.28)). Therefore, in the low-frequency range, normal seismic P- and S-waves
propagate in the undrained regime.
2.5 The quasi-static approximation of poroelasticity

The quasi-static approximation of poroelastic dynamic equations is broadly applied
for analyzing different problems of strain, stress and pore-pressure evolution in
fluid-saturated rocks, including problems related to earthquake mechanics. It was
developed in a series of now-classical works including Rice and Cleary (1976),
Rudnicki (1986), Detournay and Cheng (1993) and others. Summarizing reviews
can be found, for example, in the books of Wang (2000), Jaeger et al. (2007) and
Segall (2010). In this approximation the following system of equations (initially
containing neither sources of the fluid mass nor body forces) is considered as a
starting point. This is the continuity equation (2.64) expressing the fluid mass con-
servation, and is Darcy’s law (2.66) describing a stationary filtration of the pore
fluid. Finally, this is the equilibrium equation for the poroelastic stress:
∂
σi j = 0. (2.124)
∂x j
Taking into account the stress–strain relations (2.36) and (2.42) and substituting
Darcy’s law into the continuity equation we will immediately obtain:
∂ ∂χ ∂i j
(λdr + Mα 2 ) − αM + 2μdr = 0, (2.125)
∂ ki j ∂ ∂
(M(χ − α)) = χ . (2.126)
∂ xi η ∂ x j ∂t
This is the system of equations of the quasi-static approximation of the poroelastic-
ity. Note that equations (2.125) and (2.126) coincide exactly with equations (2.108)
and (2.117), respectively. However, the latter pair of equations describes slow
wavefields. Thus, the quasi-static approximation describes slow wavefields too.
Does the quasi-static approximation describe more phenomena than just slow
wavefields? To answer this question let us, for simplicity, consider further
elastically and hydraulically homogeneous and isotropic rocks. These assumptions
are usually accepted in the quasi-static approximation. Equation (2.126) simplifies
then to
k 2 ∂
∇ (M(χ − α)) = χ . (2.127)
η ∂t
Let us consider further the equation system (2.125) and (2.127) under undrained
conditions. Note that, in this case, the latter equation reduces to ∇ 2 = 0, which is
the rather general and well-known equation (1.50) for elastic equilibrium.
For clarity, we will denote field quantities (i.e. the pore pressure, stresses, strains,
particle velocities and displacements) in undrained conditions with a superscript u.
Then for the stress equilibration equation (2.125) we obtain
∂ u ∂iuj
(λdr + Mα 2 ) + 2μdr = 0. (2.128)
∂ xi ∂x j
This is the stress equilibrium equation with elastic parameters of an undrained
medium. Note also that the following relations are valid under undrained condi-
tions:
∇ 2 u = 0, (2.129)
χ u = 0, (2.130)
and from (2.46):
Ppu = −Bσkk
u
/3. (2.131)
Let us denote solutions of equations (2.125) and (2.127) describing slow
wavefields as
s , isj , χ s . (2.132)
Then, the following fields will also be solutions of these equations:

= u + s ,
i j = iuj + isj , (2.133)
χ =χ +χ ≡χ .
u s s
It is clear that the stress tensor and the pore pressure will also follow this rule:
σi j = σiuj + σisj ,
(2.134)
Pp = Ppu + Pps = Pps − Bσkk
u
/3.
Therefore, in addition to the dynamics of slow wavefields, the quasi-static
approximation of poroelasticity describes elastic equilibrium in undrained rocks.
However, a sufficient time interval is required for propagating normal shear and
longitudinal elastic waves to establish elastic equilibrium. Therefore, the quasi-
static approximation is not applicable for time intervals of the order of, or shorter
than, several travel times of elastic waves in systems under consideration. From the
analysis of plane-wave solutions we know that non-vanishing fluid flow in respect
to the solid skeleton is related to slow wavefields only (see Section 2.4.2). Normal
elastic waves in the quasi-static limit are characterized by vanishing relative fluid
motion (i.e. vanishing χ ). Solutions of the quasi-static equations can be represented
in the form of a summation of dynamic slow wavefields with static stress and strain
fields given by elastic equilibrium of undrained rocks. The dynamics of the slow
wavefields is mainly described by diffusion equations.
Further, we will introduce some additional useful equations of the quasi-static
approximation. We apply the divergence operator to equations (2.125):
H ∇ 2 = α M∇ 2 χ . (2.135)
This equation has several equivalent forms following from (2.37) and (2.42)–
(2.44). For example:
μdr
4α ∇ 2 Pp = −∇ 2 σii (2.136)
λdr + 2μdr
and
4μdr ∇ 2 = −∇ 2 σii (2.137)
and
4μdr α M∇ 2 χ = −H ∇ 2 σii (2.138)
and
(λdr + 2μdr )∇ 2 = α∇ 2 Pp (2.139)
and also
∇ 2 χ = S∇ 2 Pp . (2.140)
Substituting (2.135) into equation (2.127) we obtain the familiar diffusion

equation (coinciding with equation (2.111)):
∂
D∇ 2 χ = χ. (2.141)
∂t
From our discussion above it is clear that this equation (equally the quantity χ
generally) describes just slow wavefields. Thus it is not surprising that this equation
is the same as (2.111).
Using again relations (2.37), (2.42), (2.43) and (2.44) we can rewrite the
diffusion equation (2.141) in several equivalent forms. For example, using equa-
tion (2.42) we can rewrite it as follows:
∂
D∇ 2 (Pp + Mα) = (Pp + Mα). (2.142)
∂t
Another possible form is obtained from a direct substitution of equation (2.44) into
(2.141):
1 ∂ 1
D∇ 2 (K u − σii ) = (K u − σii ). (2.143)
3 ∂t 3
Also, equation (2.43) can be substituted into the diffusion equation (2.141) and we
obtain another form derived by Rice and Cleary (1976):
B ∂ B
D∇ 2 (Pp + σii ) = (Pp + σii ). (2.144)
3 ∂t 3
Note that this equation directly describes dynamics of the pressure perturbation
corrected for its undrained component (see also equations (2.134)), i.e. again, we
arrive at the same diffusion equation addressing slow wavefields. All the diffusion
equations derived above describe dynamics of the same slow wavefields. In the
case of an undrained system they turn into the identity 0 ≡ 0 (slow wavefields do
not exist).
In addition to these forms of diffusion equations there exist different mixed
forms. One can use various equivalent forms of equation (2.135). For example,
using equations (2.42) and (2.140) in equation (2.141) we obtain
kM 2 ∂ ∂
∇ Pp = Pp + α M . (2.145)
η ∂t ∂t
Also, using equation (2.37) we obtain
k B K dr 2 ∂ B
∇ Pp = (Pp + σii ). (2.146)
αη ∂t 3
Therefore, under rather general conditions (without assuming irrotational displace-
ments) the pore pressure is given by solutions of diffusion equations with coupling
dilatational strain (or equivalently, isotropic stress) terms. Such solutions describe
a complete pore pressure resulting from an interference of slow wavefields with
a pressure component given by establishing an initial elastic equilibrium in the
undrained regime. These solutions also take into account elastic fields resulting
from interactions of the slow wave with geometric boundaries of the medium. If
the medium is completely undrained then these (mixed-form) equations turn into
equation ∇ 2 Pp = 0. Under undrained conditions this is equivalent to the famil-
iar equation ∇ 2 = 0 (compare to (1.50)). Under conditions of an irrotational
displacement field and in the absence of body forces in infinite or semi-infinite
domains where and Pp vanish at infinity (see also Detournay and Cheng, 1993),
the undrained components of Pp will be equal to zero. Then the diffusion equa-
tions above will reduce to one describing slow wavefields only. The complete
consideration will be the same as we applied in the section on slow wavefields.
Equation (2.114) will then describe the diffusion of the complete pore-pressure
field.
In realistic situations of point-like borehole fluid injections (including far dis-
tances from finite-size injection sources) in large-scale reservoir domains far from
their boundaries, dynamical parts of the pore-pressure fields are given by slow
wavefields and can be well described by diffusion equations. Such linear equa-
tions were implicitly or explicitly used in many works on hydraulically induced
seismicity. Later we will also extensively use them.
2.6 Sources of fluid mass and forces in poroelastic media

In the previous sections we discussed linear differential equations of the quasi-
static approximation of poromechanics. These equations describe phenomena
relevant for understanding of fluid-induced seismicity. Here we will discuss sev-
eral useful particular solutions to these equations. Such solutions can be found if
initial and boundary conditions for the corresponding equations have been formu-
lated (see the next section). One more possibility is to directly introduce forces and
mass sources into equations and to use corresponding Green’s functions. In this
section we will introduce several solutions of the Green’s function type. We will
mainly adopt the approach of Rudnicki (1986).
We must modify the basic equations of the quasi-static approximation to intro-
duce the forces and sources of a fluid. A fluid source can be directly introduced
into continuity equation (2.64):
∂χ ∂qi
+ = Q(r, t), (2.147)
∂t ∂ xi
where Q(r, t) is a source of an additional fluid volume (per unit volume of rock
per unit time). Note that we work with linearized equations of poromechanics.
Therefore, changes of the fluid density have been neglected here. A force can be
directly introduced into the equilibrium equation for the poroelastic stress:
∂
σi j + Fi = 0, (2.148)
∂x j
where F j is a component of force density vector F.
We assume further that the medium is homogeneous and isotropic. Tak-
ing into account the stress–strain relations (2.36) and (2.41) and substituting
Darcy’s law (2.66) into the continuity equation (2.147) we obtain the following
modifications of equations (2.125) and (2.127):
∂ ∂χ ∂i j
(λdr + Mα 2 ) − αM + 2μdr + Fi = 0, (2.149)
k ∂
Q(r, t) + ∇ 2 (M(χ − α)) = χ . (2.150)
η ∂t
Analogously to equation (1.49), equation (2.149) can be written in the following
vectorial form:
∇(H − α Mχ ) − μdr ∇ × (∇ × u) + F = 0. (2.151)
Applying the divergence operator to this equation yields
∇ 2 (H − α Mχ ) + ∇F = 0. (2.152)
Expressing from here ∇ 2 and substituting it into equation (2.150) yields the
known diffusion equation (2.141) along with the source terms:
∂ αD
χ = D∇ 2 χ + Q(r, t) + ∇F. (2.153)
∂t λdr + 2μdr
The contribution of the body forces to the slow wavefield χ is non-trivial. Based
on an earlier observation of Cleary (1977), Rudnicki (1986) proposed the following
interpretation. Let us consider a fluid-mass source in the form of point-like dipoles
distributed in space and time. This is a combination of fluid sources and sinks
Q = Q 0 f (r, t)/|δr| and Q = −Q 0 f (r, t)/|δr|, respectively. Function f (r, t)

describes the spatio-temporal distribution of them. In the case of a single dipole
the source and the sink are located close to each other along vector δr. This results
in the following fluid-mass source:
Q0 ∂ f (r, t)
lim ˆk
( f (r − δr, t) − f (r, t)) = −Q 0 δr , (2.154)
|δr|→0 |δr| ∂ xk
ˆ k is the k-component of the unit vector δr
where δr ˆ ≡ δr/|δr|. This source will
produce exactly the same effect as the following body force:
ˆ λdr + 2μdr
F = −Q 0 δr f (r, t). (2.155)
αD
Or equivalently, the fluid-mass effect of the body-force distribution F0 f (r, t) is
equal to the effect of the fluid-mass dipole with:
ˆ = −F0 αD
Q 0 δr . (2.156)
λdr + 2μdr
Thus, the following strategy can be applied to solve the system of poroelastic
equations with source terms, i.e. equations (2.151) and (2.153). The case of a
point-like fluid-mass source corresponds to the absence of any body forces, i.e.
F = 0 (recall that we neglect effects of gravitation forces). Such a situation approx-
imately corresponds to a point-like fluid injection from a borehole perforation into
rocks. In this case firstly a solution of equation (2.153) for the slow wavefield
χ must be obtained. Owing to the symmetry of the problem such a solution of
equation (2.151) for the displacement field u is completely irrotational. All the
field quantities describe slow wavefields only. Owing to the quasistatic approxi-
mation, normal seismic waves are not described. Their contribution to any elastic
equilibrium is assumed to occur infinitely quickly.
Very similar to the previous one is the case of a fluid-mass dipole only. Again,
F = 0. This is a superposition of two point-like positive and negative fluid-mass
sources. Such a situation approximately corresponds to a pair of closely located
boreholes with a point-like fluid injection in one of them and a point-like fluid
extraction in the other (e.g. a geothermal duplet considered at distances much larger
than the distance between the boreholes, i.e. in the far field). Thus, a solution for the
slow wavefield χ and for other field quantities can be found as a superposition of
the corresponding solutions for a single fluid-mass source. Taking into account that
a dipole source is equal to two differentially closely located mass sources of oppo-
site signs, one can show that such a superposition is proportional to a directional
derivative of of the solutions for a single mass source (see Rudnicki, 1986). The
field quantities describe slow wavefields only. They are completely irrotational and
satisfy (2.151) with zero body forces. Normal seismic waves are not considered.
The corresponding elastic part of the equilibrium is assumed to establish itself

instantaneously.
In both situations discussed above, slow wavefields fully represent the dynamics
of systems and, thus, corresponding solutions satisfy equations (2.112)–(2.115).
The case of an applied body force (F = 0) and a vanishing source of the fluid
mass (Q = 0) is different. A solution for the slow wavefield χ can be first found for
a fluid-mass dipole (2.156). This solution corresponds to an irrotational part of the
displacement field satisfying (2.151) with zero body forces. This solution describes
the slow wavefield. The complete solution will be the sum of the slow-wavefield
solution and the solution of equation (2.151) for χ = 0 and the force F. This latter
solution corresponds to the zero slow wavefield. It describes elastic equilibrium
established in the undrained rock by normal seismic waves (again, these waves
are not considered in the quasi-static approximation; therefore, this approximation
assumes that the undrained elastic equilibrium is established momentarily).
Other solutions of quasi-static poroelastic equations with source terms can be
obtained by superposition of the solutions discussed above for the cases with load-
ing of the form Q = 0 and F = 0 and with loading Q = 0 and F = 0. A
rather general form of the superposition principle (1.37) for a linear equation sys-
tem (including the linear equation system of poroelasticity considered above) uses
Green’s functions. A review of poroelastic Green’s functions valid in the total
frequency range was given by Karpfinger et al. (2009). Numeric inverse Fourier
transforms are required for their computations. In this book we are interested in the
low-frequency (quasi-static) range. In this range several explicit Green-function-
like solutions are known (see Rudnicki, 1986). However, quasi-static poroelastic
equations do not describe dynamic effects on small time scales (time intervals on
the order of normal seismic travel times). Thus, in this approximation it is conve-
nient to work with source terms having Heaviside-type temporal distributions h(t)
(h(t) = 0 if t < 0 and h(t) = 1 if t ≥ 0).
2.6.1 Fluid injection at a point of a poroelastic continuum

Let us consider a point-like source of an instantaneous fluid injection of a volume
Q 0 at the time moment t = 0 located at the point r = 0. The source function has
the following form:
Q(r, t) = Q 0 δ(r)δ(t). (2.157)
In this case F = 0 and, as discussed above, all field quantities represent a slow
wavefield. Thus we must find a solution for the quantity χ . Therefore, we consider
diffusion equation (2.153) in an infinite poroelastic 3D medium with χ = 0 at
t < 0:
∂
χ = D∇ 2 χ + Q 0 δ(r)δ(t). (2.158)
∂t
Solution χG of this equation is Green’s function of the diffusion equation (2.153).
This classical solution can be found by reformulating the problem into an initial-
value one. A spatial Fourier transform is used to turn the homogeneous diffusion
equation into an ordinary one-dimensional (1D) differential equation (see the
derivation in paragraph 51, chapter V, of Landau and Lifshitz, 1991):

Q0 r2
χG (r, t) = exp − . (2.159)
(4π Dt)3/2 4Dt
We see that, because of the spherical symmetry of the problem, χG is a function

of the distance r = |r| and time t only (no angular dependencies). Owing to the
diffusional dynamics the normalized distance
r √ r
r0 = √ = π (2.160)
4Dt λP I I
is a convenient variable to describe spatio-temporal behavior of the field quantities.

The right-hand part of this equation uses the wavelength of the slow wave with the
time period t (see equation (2.104)).
In reality an instantaneous fluid injection is a mathematical abstraction. The
following situation is more relevant in practice. We consider further a point-like
source of a continuous fluid injection of a constant volume rate Q I located at
r0 = 0 and starting at t = 0 (i.e. the injection rate is the following func-
tion of time: Q I h(t)). The source function is given by the time integration of
this rate:
Q(r, t) = Q I tδ(r). (2.161)
Also in this case F = 0 and, as discussed above, all field quantities represent a
slow wavefield. Owing to the symmetry of the problem the displacement vector
u will have a radial direction. Scalar-dilatational and pressure-like field quantities
, χ, Pp and σii will be functions of the distance r and time t only (no angular
dependencies).
The solution of diffusion equation (2.153) in this case is very well known (it is
given by an integration over time of Green’s function (2.159); see, for example,
Carslaw and Jaeger, 1973):
QI
χ (r, t) = erfc(r0 ), (2.162)
4π Dr
1
0.8
0.6
0.4
0.2
0
0 0.5 1 1.5 2
x
erf(x) erfc(x) exp(–x)
Figure 2.6 An error function, a complimentary error function and a decreasing

exponential function.
where erfc(x) stands for the complementary error function, erfc(x) = 1 − erf(x),
and the error function erf(x) is defined as follows (see also Figure 2.6):
x
2
erf(x) ≡ √ exp(−z 2 )dz. (2.163)
π 0
Consequently, the pore pressure is given by equation (2.113):

QI η
Pp (r, t) = erfc(r0 ). (2.164)
4πkr
Using equations (2.162) and (2.151) and taking into account radial symmetry,
irrotationality and physical dimensionality of the displacement, Rudnicki (1986)
derives the following result:
Q I ηxi n s
u i (r, t) = srr (r0 ), (2.165)
16π kr μdr
where the function srr (r0 ) is defined as follows:
erf(r0 ) exp(−r02 )
srr (r0 ) ≡ 2 erfc(r0 ) + − 2 √ , (2.166)
r02 r0 π
and we have introduced the poroelastic stress coefficient n s given by
α(1 − 2ν dr ) αμdr SM
ns ≡ = = αμdr . (2.167)
2(1 − ν dr ) λdr + 2μdr H
Here ν dr is Poisson’s ratio of the drained rock. Substituting result (2.165) into
Hooke’s law (2.36) yields the stress tensor (Rudnicki, 1986):
QI η xi x j
σi j (r, t) = −n s δi j stt (r0 ) + 2 sdi f (r0 ) , (2.168)
4πkr r
where the functions stt (r0 ) and sdi f (r0 ) are defined as follows:
stt (r0 ) ≡ erfc(r0 ) − + √ , (2.169)
2r02 r0 π
sdi f (r0 ) ≡ srr − stt = erfc(r0 ) + 3 − 3 √ . (2.170)
2r02 r0 π
Stress tensor (2.168) corresponds to the spherical slow wavefield radiated by the
point-like mass source considered above.
One can write these results in terms of the radial and tangential components of
the corresponding field quantities. For the radial component of the displacement
(its tangential components vanish) we have
Q I η ns
u r (r, t) = srr (r0 ). (2.171)
16π k μdr
For the radial component of the stress σrr (it is equal, e.g. σ11 at r = (x1 , 0, 0)) we
obtain
QI η
σrr (r, t) = −n s srr (r0 ). (2.172)
4πkr
For the tangential components σθθ = σφφ (they are equal, e.g. σ33 at r = (x1 , 0, 0))
we obtain
QIη
σθθ (r, t) = −n s stt (r0 ). (2.173)
4πkr
Rudnicki (1986) also found explicit analytical expressions of Green’s functions
for source terms like a fluid-mass dipole, a point force, a line source of a fluid mass,
a line of fluid-mass dipoles, and a line force.
2.7 Boundary loading of poroelastic media

Let us consider a loading of a poroelastic medium over its boundaries. Detournay
and Cheng (1993) proposed decomposing the problem into two loading modes.
Mode I denotes the following load distribution: to a given surface a normal stress
σn (we recall that this is a normal confining, i.e. a normal total stress) is applied
so that σn (t) = −h(t) p0 and the applied pore pressure is absent, Pp = 0. Mode II
denotes the following loading: σn = 0 and Pp = h(t) p0 . A corresponding super-
position of solutions of quasi-static poroelastic equations for these two modes will
provide solutions for arbitrary distributions of pore pressure and normal stress at
the boundary.
2.7.1 Loading of a poroelastic half-space by a fluid reservoir

Let us use this strategy to compute poroelastic field perturbations excited by a
sudden application of fluid pressure p0 to the surface of a poroelastic half-space.
Clearly this is an example of a 1D problem. All quantities are functions of one
variable (e.g. of depth z) only. The boundary conditions at the surface of the half-
space z = 0 are σzz = −h(t) p0 and Pp = h(t) p0 . It is clear also that there
is only one non-vanishing strain component zz (because no displacements along
horizontal directions are possible). Thus, this problem is also frequently called a
uniaxial-strain problem.
For such a uniaxial-strain problem the stress–strain relation (2.36) takes the
following simplified forms:
σzz = (λdr + 2μdr )zz − α Pp (2.174)
and
σx x = σ yy = λdr zz − α Pp . (2.175)
The pore-pressure relation (2.42) takes the following form:
Pp = M(χ − αzz ). (2.176)
Also we will need the equilibrium equation for the poroelastic stress (2.148). Its
correspondingly simplified form is
∂
σzz = 0. (2.177)
∂z
Further, it is convenient to use and simplify the diffusion equation for the slow
wavefield in the mixed form (2.145):
k M ∂2 ∂ ∂
Pp = Pp + α M zz . (2.178)
η ∂z 2 ∂t ∂t
Using the stress–strain relation (2.174) we obtain:
∂2 ∂ αM ∂
D Pp = Pp + σzz . (2.179)
∂z 2 ∂t H ∂t
This equation coincides with equation (141) of Detournay and Cheng (1993).
Further, we will denote field quantities resulting from the loading modes I and II
by the superscripts I and II, respectively. In loading mode I the stress applied to
the surface of the half-space is σzzI (0, t) = − h(t) p0 . Simultaneously, PpI (0, t) = 0.
We must now find field quantities satisfying these boundary conditions and
equations (2.174)–(2.179).
From equation (2.177) it follows that the total stress is independent of z. Thus,
σzzI (z, t) = −h(t) p0 . (2.180)
Therefore, for times t > 0 this stress is constant in time too. Thus, equation (2.179)
takes a form of a 1D diffusion equation for the pore pressure:
∂2 I ∂
D Pp = PpI . (2.181)
∂z 2 ∂t
It is clear that in the first instant after the load application the system was in
the undrained state (i.e. at this moment the fluid flow was vanishing: χ I (z, 0) = 0).
Correspondingly with equation (2.176) the pore pressure under undrained condi-
tions is directly related to the uniaxial strain:
PpI (z, 0) = −Mαzz

I
. (2.182)
Under undrained conditions using this relation in equation (2.174) and using defi-
nition (2.167) we can further directly relate the pore pressure to the loading stress:
ns I
PpI (z, 0) = − σ (z, 0), (2.183)
Sμdr zz
Therefore, in the first instant the load should lead to a step-like increase of the
pore pressure equal to n s p0 h(t)/(Sμdr ) everywhere in the medium (also at infin-
ity). On the other hand, in the loading mode I the boundary condition for the pore
pressure is PpI (0, t) = 0. The solution of equation (2.181) satisfying these two
facts (we have already seen that the erfc-function is a solution of a homogeneous
1D diffusion equation) is:
n s p0 z
PpI (z, t) = (h(t) − erfc( √ )). (2.184)
Sμdr 4Dt
In the loading mode II σzzII (z, t) = 0 and PpII (0, t) = p0 h(t). The corresponding
solution of equation (2.181) is
z
PpII (z, t) = p0 erfc( √ ). (2.185)
4Dt
Note that this expression will not change after multiplication by h(t). Therefore,
the complete solution of the boundary-value problem is the sum of the solutions
for the modes I and II. It has the following form:

ns ns z
Pp (z, t) = p0 h(t) + (1 − ) erfc( √ ) . (2.186)
Sμdr Sμdr 4Dt
This form of the pore-pressure distribution for a poroelastic half-space loaded by

a water reservoir was first obtained by Roeloffs (1988). In this classical paper
Roeloffs (1988) also obtained a semi-analytical semi-numerical solution for a
poroelastic half-space loaded by a finite-width fluid reservoir (in the plane-strain
geometry) located on the half-space surface. She applied her solution to ana-
lyze reservoir-induced seismicity. These poroelastic solutions and several related
approximations were used in an extensive series of interpretation and modeling
studies of induced seismicity at the Koyna–Warna artificial water reservoirs (see
Talwani, 1997; Gupta, 2002; Kalpna and Chander, 2000; Gavrilenko et al., 2010;
Guha, 2000; and further references therein).
Using (2.186) and (2.174)–(2.175) and taking into account that σzz (z, t) =
σzzI (z, t) is given by (2.180) we obtain:

p0 h(t) ns ns z
zz (z, t) = − 1−α + 1− erfc √
λdr + 2μdr Sμdr Sμdr 4Dt
(2.187)
and the lateral components of the stress:

λdr ns ns z
σx x = σ yy = − p0 h(t) + 2n s + 1− erfc √ .
λdr + 2μdr Sμdr Sμdr 4Dt
(2.188)
Finally, the fluid-filtration caused change of the pore-space volume per unit rock
volume, χ(z, t), can be computed by using equation (2.176):

ns z
χ (z, t) = p0 h(t)S 1 − erfc √ . (2.189)
Sμdr 4Dt
2.7.2 Fluid injection into a spherical cavity of a poroelastic continuum

In this book we are interested in fluid injections in rocks over surfaces of cavities.
For example, a fluid injection into an open short borehole section is used for devel-
opment of enhanced geothermal systems. Very roughly, one can consider such an
injection as an application of a fluid pressure to the surface of a spherical cavity.
Let us consider a spherical cavity of radius a0 located at the origin of a fluid-
saturated porous elastic infinite continuum with initially absent stress and pressure
perturbations. We assume a sudden application of a fluid pressure h(t) p0 to the
inner surface of the cavity. This loading is equal to a direct superposition of modes
I and II described above. For r > a0 we must solve the system of poroelastic
equations, i.e. equations (2.151) and (2.153) without source terms.
First we consider the loading mode I. In this mode no sources of fluids are
present in the medium. The normal stress applied to the surface of the cavity is
σrrI (a0 , t) = −h(t) p0 . A pore-pressure perturbation vanishes at both limits, at the
cavity surface as well as at infinity: PpI (a0 , t) = PpI (∞, t) = 0. Thus, no slow
wavefields are being excited. Therefore, we must consider elastic deformations of
the undrained rock. The solution of such a boundary-value problem is well known
(Landau and Lifshitz, 1987, paragraph 7, problem 2). There is a radial displacement
component only:
p0 a03 xi
u iI = h(t) . (2.190)
2μdr r 3
This can be reformulated for the radial displacement directly:
p0 a03
u rI = h(t) . (2.191)
2μdr r 2
This also yields I = ∇u = 0 and σkk
I
= 0. Therefore, no pore-pressure per-
turbations have been induced by this undrained deformation. For the radial and
tangential strains one obtains:
p0 a03
rr
I
= −h(t) , (2.192)
μdr r 3
p0 a03
θθ
I
= φφ
I
= h(t) . (2.193)
2μdr r 3
The radial and tangential stresses are:
a03
σrrI = −h(t) p0 , (2.194)
r3
a03
σθθ
I
= σφφ
I
= h(t) p0 . (2.195)
2r 3
Let us further consider loading mode II. The normal stress applied to the surface
of the cavity is vanishing, σrrII (a0 , t) = 0. A pore-pressure perturbation vanishes at
infinity. However, at the cavity surface we have PpII (a0 , t) = p0 h(t). Owing to the
symmetry of the problem it is clear that only irrotational wavefields can be excited.
Because of the nature of the boundary problem it is convenient to work directly
with the diffusion equation for the pore pressure instead of equation (2.153).
The corresponding initial- and boundary-value problem for the diffusion equa-
tion (2.114) is very well known. For example, it can be solved in the same way
as the problem 1 of paragraph 52, chapter V of Landau and Lifshitz (1991):
a 0 p0
PpII (r, t) = erfc(ra ), (2.196)
r
where
r − a0 √ r − a0
ra = √ = π (2.197)
4Dt λ PII
is the normalized distance r0 counted from the cavity surface. By this loading mode
only slow wavefields can be radiated (i.e. slow wavefields define completely the
field quantities). Thus, we use equation (2.113) to obtain the field χ :
a 0 p0 S
χ II (r, t) = erfc(ra ). (2.198)
r
Results (2.196) and (2.198) will coincide with equations (2.164) and (2.162),
respectively, if we make the following substitutions:
ra → r0 (2.199)
QI η
a 0 p0 → . (2.200)
4πk
Furthermore, because only slow wavefields are excited, we can directly apply
equations (2.112) and (2.115) to compute the corresponding dilatational field
quantities, respectively:
a 0 p0 n s
II (r, t) = erfc(ra ), (2.201)
μdr r
4a0 p0 n s
σiiII (r, t) = − erfc(ra ). (2.202)
r
On the other hand, the dilatation II is given by the divergence of the dis-
placement vector field. Owing to the spherical symmetry of the problem the
displacement can have a radial component only. Therefore, in the spherical
coordinate system with the origin at the cavity’s center, we have
1 ∂r 2 u rII
II (r, t) = ∇uII = . (2.203)
r 2 ∂r
Using this equation and equation (2.201) we obtain the displacement by the
integration of the dilatation over r :

(r − a0 )2 a0 p0 n s 1 a0 CII
u r (r, t) =
II
srr (ra ) + s (ra ) + 2 , (2.204)
μdr r 2 4 r − a0 r
where we have introduced the following new notation
1
s (ra ) ≡ erfc(ra ) + √ (1 − exp(−ra2 )), (2.205)
ra π
and CII is an integration constant. This constant must be found from the boundary
conditions. Later we will see that CII = 0.
Because of the radial symmetry the components of the displacement vector in a
Cartesian coordinate system with the origin at the cavity’s center are
xi
u iII (r, t) = u rII (r, t) . (2.206)
r
From here we compute the strain rr II

, which is equal to xIIx at points (x, 0, 0):

a0 p0 n s erfc(ra ) (r − a0 )2 1 a0 CII
rr =
II
−2 ( srr (ra ) + s (ra )) − 2 3 .
μdr r r 3 4 r − a0 r
(2.207)
To compute the radial-stress component it is convenient to directly apply
equation (2.36). The stress σrrII is equal to σxIIx at points (x, 0, 0):
σrrII = λdr II + 2μdr rr
II
− α Pp . (2.208)
Using here results (2.201) and (2.207) and taking into account the boundary
condition of loading mode II that σrrII (a, t) = 0 (this yields CII = 0) we obtain:

(r − a0 )2 a0 p0 n s 4a0
σrr (r, t) = −
II
srr (ra ) + s (ra ) . (2.209)
r3 r − a0
Further, the twofold tangential-stress component can be obtained by extraction of
this result from equation (2.202). Because σθθ II
= σφφ
II
we obtain

a 0 p0 n s (r − a0 )2 2a0 (r − a0 )
σφφ = σθθ = −
II II
2 erfc(ra ) − srr (ra ) − s (ra ) .
r 2r 2 r2
(2.210)
The solution of the complete boundary-value problem for the corresponding field
quantities is given by the following sums:
u = uI + uII , (2.211)
where (2.191) and (2.204) must be substituted (note again that, in (2.204), CII = 0).
Further,
σ = σ I + σ II , (2.212)
where (2.194), (2.195), (2.209) and (2.210) must be substituted, respectively.
Finally, the following two quantities are given directly by
χ = χ II , Pp = PpII , (2.213)
where (2.198) and (2.196) must be substituted, respectively.
However, at distances r a0 (i.e. in the far field of the injection cavity) one
can easily see that the contributions uI and σ I become negligible compared to con-
tributions uII and σ II , respectively. We note also that, at large distances, ra can be
replaced by r0 . Therefore, the solution for the boundary-value problem considered
here nearly coincides with the solution for the mass source from Section 2.6.1. In
the far field their only difference is a constant factor describing the strength of the
injection source. In the case of a fluid-mass source this factor is Q I η/(4πk). In the
case of a pressure applied to the surface of an injection cavity this factor is a0 p0 .
Using this fact along with the symmetry considerations we conclude that in
far field of the cavity the corresponding solution of equation (2.151) for the
displacement and stress fields can be obtained by the direct modification of

equations (2.165) and (2.168), respectively:
a 0 p0 x i n s
u i (r, t) = srr (r0 ), (2.214)
4r μdr
and
a 0 p0 xi x j
σi j (r, t) = −n s δi j stt (r0 ) + 2 sdi f (r0 ) . (2.215)
r r
Again, we write these results in terms of the radial and tangential components
of the corresponding field quantities. For the radial component of the displacement
(its tangential components are vanishing) we have
a 0 p0 n s
u r (r, t) = srr (r0 ). (2.216)
4 μdr
For the radial component of the stress σrr we obtain
a 0 p0
σrr (r, t) = −n s srr (r0 ). (2.217)
r
For the tangential components we obtain
a 0 p0
σφφ (r, t) = σθθ (r, t) = −n s stt (r0 ). (2.218)
r
Concluding this section, we note that the direct applicability of the homo-
geneous diffusion equation (2.114) to the pore pressure in the boundary-value
problem considered here is a consequence of the following circumstances. First,
the symmetry of the problem allows for an irrotational displacement field only.
This corresponds to the exclusive radiation of the slow wavefield in loading mode
II. Second, and more importantly, we have considered here propagation of pore-
pressure perturbations in an infinite spatial domain. In spatially bounded bodies,
even under conditions of irrotational displacements, the homogeneous diffusion
equation for the pore pressure (2.114) is not directly applicable. The physical rea-
son of this is the fact that slow wavefields interact with boundaries (surfaces or
interfaces). As a result of such interactions at the boundaries the slow wavefields
would be reflected into secondary slow wavefields and transformed into normal
elastic waves (converted P-waves). Pore-pressure perturbations corresponding to
the secondary slow wavefields will still satisfy equation (2.114) directly. How-
ever, pressure-perturbation components corresponding to the normal elastic waves
(propagating infinitely quickly in the quasi-static poroelastic approximation) will
not satisfy this equation.
The well-known (in the quasi-static approximation) Mandel–Cryer effect (see
Detournay and Cheng, 1993; Jaeger et al., 2007; and references therein) can be also
understood in terms of conversion (reflection) of a normal P-wave into the slow
wave at the boundaries of finite poroelastic domains. Indeed, the Mandel–Cryer
effect is clearly observed in the following situations. An instantaneous pressure
pulse p0 δ(t) or a constant normal stress p0 h(t) is instantly applied to the surface
of a fluid-saturated porous elastic spherical or cylindrical sample with radius a0

and a vanishing initial pore pressure. In the second situation, the sample is sub-
jected to loading mode I discussed above. The pore pressure in the center of the
sample will instantaneously increase up to the level corresponding to the undrained
sample (note that in the quasi-static approximation this jump in the pore pressure
occurs instantaneously, because normal elastic waves propagate in this approxima-
tion with infinite velocity). Then, the pore pressure will start to decrease slowly
(due to the fluid filtration). However, in contrast to the expectation of a simple dif-
fusion approximation for the pore pressure (neglecting the poroelastic coupling of
the pore pressure and the stress in a finite spatial domain), the temporal evolution
of the pore pressure is not monotonic. First the pore pressure in the center of the
sample will rise slightly above the undrained level instantaneously reached at the
time t = 0 and only after some time will it start to slowly decrease.
Detournay and Cheng (1993) show an example of numerical simulation of the
loading of an infinitely long cylinder sample with the undrained and drained Pois-
son’s ratios νu = 0.15 and νdr = 0.31 (see their figure 3). The pore pressure in the
center of the sample exceeds the undrained level by about 10% at most. According
to the standard explanation this occurs because the filtration of the fluid from the
surface vicinity of the sample leads to softening of this vicinity. This vicinity of
the sample surface can be compressed even more. This produces an additional load
to the center of the sample and leads to an additional increase of the pore pressure
there.
Alternatively one can look at this effect from the point of view of wave propaga-
tion. Instantaneous loading of the surface of the sample produces normal P-waves
propagating in the sample. These waves converge in the center of the sample and
spread further towards the sample’s surface. The normal P-waves quickly establish
the undrained pore-pressure level. On the other hand, they interact again with the
sample surface and a part of their energy becomes converted into slow P-waves.
These waves correspond to the fluid filtration in the sample and from the sample
outwards. On the surface of the sample the slow waves are also converted into
the normal P-waves responsible for adjusting the elastic equilibrium. Slow waves
require more time for propagation in the sample. This time delay leads to a slow
rise of the pore pressure in the sample’s center above the undrained level. Interest-
ingly, from figure 3 from Detournay and Cheng (1993), one can estimate that the
time of the maximum pore pressure in the center corresponds approximately to the
propagation time required by the slow wave of the sample-radius length to reach
the center.
Detournay and Cheng (1993) give analytical solutions of several poroelastic
boundary-value problems in finite spatial domains (e.g. a poroelastic cylinder
whose surface has been suddenly exposed to a fluid pressure). They show that
2.8 Stress and pressure coupling for radially symmetric fluid sources 95
direct application of equation (2.114) does not provide the exact solution of the
problem. Under such circumstances the homogeneous diffusion equation must first
be solved for the quantity χ .
Descriptions of many boundary-value problems of linear quasi-static poroelas-
ticity and their analytical solutions can be also found in the book by Wang (2000).
2.8 Stress and pressure coupling for radially symmetric fluid sources
Poroelastic coupling can affect seismicity induced by borehole-fluid injections
and extractions. In this section we consider effects of poroelastic coupling onto
a change FC S of the failure criterion (1.73). For this we use the exact solutions
by Rudnicki (1986) for a point-like fluid-mass source in a poroelastic medium pre-
sented in the previous section, (2.164), (2.172) and (2.173), and far-field limits of
solutions for a pressure-cavity source, (2.196), (2.217) and (2.218).
Poroelastic contributions to the tectonic stress as functions of time t and distance
r are given by the pore-pressure perturbation Pp and by radial and tangential stress
perturbations. The pore-pressure perturbation is counted relative to the natural level
of the in situ pore pressure in rocks (which is commonly, but not always, close to
the hydrostatic one). In spherical coordinates the radial component of the stress per-
turbation is equal to the radial component of the poroelastic traction perturbation
acting on a plane with a normal parallel to the radius vector of a given location. The
tangential stress perturbation is equal to the normal component of the additional
(caused by poroelastic coupling) traction acting on a plane that includes the radius
vector of a given location. The radial poroelastic-perturbation-stress component is
σrr . Because of the spherical symmetry the tangential component is equal to σφφ =
σθθ . Figure 2.7 schematically shows the poroelastic contributions at two points of
a fluid-saturated tectonically compressed medium in the case of fluid injection.
In a spherical coordinate system with the origin coinciding with the injection
source, the poroelastic contributions (see also definitions (2.166) and (2.169)) are:
stt (r0 ) srr (r0 ) erfc(r0 )
σθθ = −n s C f , σrr = −n s C f , Pp = C f . (2.219)
r r r
The source term C f is given by
QIη
Cf = , (2.220)
4πk
where Q I is a constant injection volume rate (mass rate divided by an initial density
of the injection fluid, which is assumed be the same as the fluid saturating the
medium). For an injection with a constant pressure ( p0 ) from a spherical cavity of
the radius a0 the source term will be
p σrr σh
σθθ
σv
σθθ
σrr
σh
σv
Figure 2.7 A sketch of poroelastic-coupling-related tractions in two points of a

tectonic domain in the case of fluid injection. The injection source is symbolized
by a dark irregular spot indicated with p. We consider two points on an arbitrary
spherical surface in the medium. One of the points is at the bottom of the sphere
and another one is on the right. The crossed black lines indicate horizontal and
vertical planes at these points. Small arrows indicate tractions normal to these
planes. These are poroelastic-coupling-caused tractions σrr and σθθ . Note that
their directions depend on the point location on the sphere. However, directions
of the tectonic tractions σv and σh are independent of the location on the sphere
(we assume a uniformly distributed tectonic stress tensor). The complete stress
tensor is equal to the sum of the tectonic and poroelastic parts. At the bottom of
the sphere the resulting stress components will be σv + σrr and σh + σθθ . At the
right the stress components are σv + σθθ and σh + σrr . Therefore, even at equal
distances from the injection source poroelastic coupling leads to a heterogeneous
distribution of the resulting stress tensor.
C f = p0 a 0 . (2.221)
Figure 2.8 shows source-normalized poroelastic stresses and pore pressure cor-
rected for geometrical spreading (or spherical divergence) given by the functions
stt (r0 ), srr (r0 ) and erfc(r0 ), respectively. Let us first assume for simplicity that the
maximum compressive tectonic stress is vertical. This is a normal-faulting (exten-
sional) tectonic regime. We will then denote the maximum compressive stress as
σv . The corresponding minimum compressive stress is a horizontal one, σh . Note
that the values of these stress components are negative. An extension environment
and the fact that the maximum compressive stress is vertical mean that (recall the
notations introduced in Section 1.2):
σ1 = −σv > σ3 = −σh . (2.222)
Let us consider poroelastic-coupling contributions to the stress tensor along the

directions of the maximum compressive stress and the minimum compressive stress
i.e. along the directions corresponding to σ1 and σ3 , respectively.
First we consider points located exactly vertically above (or below) the injection
source (Figure 2.7). These are locations whose radius vectors r are parallel to
the axis xv of the maximum compression stress (with the absolute value σ1 , in
this case). Thus, the principal stresses in this location will be modified by the
poroelastic coupling as follows:
σmodh = σh + σθθ , σmodv = σv + σrr , (2.223)
where we denote by σmodv and σmodh the poroelastic-coupling-modified stresses σv

and σh , respectively. In turn, the values of principal stresses σ1 and σ3 will also be
modified. We denote them as σmod1 and σmod3 , respectively. Note that we assume
that absolute values of the tectonic stresses are significantly larger than absolute
values of the poroelastic contributions, and that poroelastic effects do not change
the order of these stresses in condition (1.52). Then (see also equation (2.222)):
σmod3 = σ3 − σθθ , σmod1 = σ1 − σrr . (2.224)
Therefore, the differential and mean stresses at such locations will be also
modified, respectively:
σmod1 − σmod3 = σ1 − σ3 + σθθ − σrr , (2.225)
(σmod1 + σmod3 )/2 = (σ1 + σ3 )/2 − (σθθ + σrr )/2. (2.226)
In these two equations the second terms on their right-hand sides give the
poroelastic-coupling contributions (i.e. changes of the differential and mean
stresses, τv and σv , respectively). Using equations (2.219) we obtain for these
coupling contributions, respectively (see also (2.170)):
τv = σθθ − σrr = n s C f sdi f (r0 )/r, (2.227)
σv = −(σθθ + σrr )/2 = n s C f sm (r0 )/r, (2.228)
where we have introduced
sm (r0 ) ≡ (stt (r0 ) + srr (r0 ))/2. (2.229)
Analogously, at points located exactly horizontally to the left and right from the
injection source (see Figure 2.7), i.e. points with radius vectors parallel to axis xh
of the minimum-compression in situ stress:
σmodh = σh + σrr , σmodv = σv + σθθ . (2.230)

In terms of the quantities σ1 and σ3 , this means:
σmod3 = σ3 − σrr , σmod1 = σ1 − σθθ . (2.231)
Correspondingly, the differential and mean stresses at such locations will also be
modified:
σmod1 − σmod3 = σ1 − σ3 + σrr − σθθ , (2.232)
(σmod1 + σmod3 )/2 = (σ1 + σ3 )/2 − (σθθ + σrr )/2. (2.233)
Again, in these two equations the second terms on their right-hand sides give
the poroelastic-coupling contributions. Using equations (2.219) we obtain these
coupling contributions:
τh = −(σθθ − σrr ) = −n s C f sdi f (r0 )/r, (2.234)
σh = −(σθθ + σrr )/2 = n s C f sm (r0 )/r. (2.235)
At all other points σmod1 and σmod3 will have intermediate values between those
given by (2.224) and (2.231). The values of principal stress σ2 will be modified in
a similar manner. It will be σ2 − σrr in points located along the acting direction
of this stress, and σ2 − σθθ in points located on the plane normal to this direc-
tion. Therefore, if the difference |σrr − σθθ | is greater than σ2 − σ3 or σ1 − σ2
then the poroelastic coupling can change the stress regime. In practice this would
require nearly vanishing differences between the principal stresses at least in one
of principal-stress planes.
Figure 2.8 shows that normalized poroelastic stress contributions sdi f and sm are
positive. Also the source term C f is positive for the case of fluid injection. For
example, it is equal to p0 a0 for a constant pressure p0 switched on at time t = 0 on
a surface of a spherical cavity of radius a0 . The radius of Mohr’s circle increases
with increasing τ . Thus, one can see from (2.227) and (2.234) that during an
injection the maximum increase of the radius of the Mohr’s circle will be at the top
and bottom locations. Its maximum decrease will be along the horizontal axis.
For fluid extractions the source term C f is negative. Thus, from (2.227)
and (2.234) we conclude that during fluid production the maximum increase of
Mohr’s circle will be along the horizontal direction, and its maximum decrease
will be along the vertical direction.
A change of the differential stress (and, therefore, radius of Mohr’s circle) is only
one of possible contributions to changes of the failure criterion stress FCS. Because
of fluid injection/extraction, the following changes of failure criterion stress (1.73)
2
1.5
0.5
0
0.01 0.05 0.1 0.5 1 5 10
r0
Pp(r0) sdif(r0) sm(r0) srr(r0)
stt(r0)
Figure 2.8 Dimensionless functions (from top to bottom at the vicinity of r0 =

0.01) srr (r0 ), sm (r0 ), sdi f (r0 ), erfc(r0 ) (it corresponds
√ to Pp in equation 2.219),
and stt (r0 ) of the dimensionless distance r0 = r/ 4Dt. These functions show
the pore pressure and poroelastic-stress perturbations without influences of the
source, of the geometrical √spreading and of the poroelastic stress coefficient. The
vertical line shows r0 = π being an approximation of the microseismic trigger-
ing front (see the next Chapter and equation 3.6) used in this book for the case of
a linear-diffusion induced seismicity.
are caused along the maximum compressive-stress direction (i.e. the vertical axis
for the normal-stress regime):

1
FC Sv = C f n s sdi f (r0 ) − (n s sm (r0 ) − erfc(r0 )) sin φ f /r. (2.236)
2
Along the minimum compressive-stress direction (i.e. the horizontal axis for the
normal-stress regime) these changes are

1
FC Sh = C f − n s sdi f (r0 ) − (n s sm (r0 ) − erfc(r0 )) sin φ f /r. (2.237)
2
Equations (2.236) and (2.237) show that during fluid injection (a positive C f )
the maximum change of FC S is FC Smax = FC Sv and its minimum change
is FC Smin = FC Sh . During fluid extraction (a negative C f ) FC Smax =
FC Sh , and its minimum change is FC Smin = FC Sv .
The poroelastic coupling is given in (2.236) and (2.237) by terms with the fac-
tor n s . Therefore, in addition to the radially symmetric destabilizing effect of the
increasing pore pressure, the poroelastic-coupling effect of fluid injection enhances
this destabilization below and above the injection source, i.e. along the direction
of the maximum in situ compressive stress. The coupling effect becomes stabiliz-
ing along the minimum in situ compressive stress (i.e. it patially compensates the
destabilizing effect of the increasing pore pressure in such locations). In the case of
fluid extraction, and under conditions of a sufficiently strong poroelastic coupling
(i.e. sufficiently high values of parameter n s ), destabilization of rocks is also pos-
sible. This effect reaches its maximum in the horizontal direction. In other words,
the destabilizing effect of fluid extraction is strongest along the minimum in situ
compressive stress direction.
These conclusions can be reformulated in the following way. In the case of an
extensional tectonic stress regime, the maximum compressive stress is vertical.
This stress regime leads to normal faulting. As a result of poroelastic coupling,
fluid injection will stimulate normal faulting, especially strongly above and below
the injection source. In contrast, fluid extraction will stimulate normal faulting,
especially strongly on a horizontal plane on flanks of the fluid-sink position.
In the case of compressional tectonic-stress regime, the maximum compressive
stress is horizontal. This leads to reverse faulting. Fluid injection will stimulate
reverse (thrust) faulting especially strongly on a horizontal plane on flanks of the
fluid-source position. In contrast, fluid extraction will stimulate reverse faulting,
especially strongly above and below the fluid-sink position.
Similar observations and conclusions were made by Segall (1989) and Segall
and Fitzgerald (1998), who computed strains and stresses for a finite poroelastic
reservoir imbedded into an elastic continuum.
The difference in poroelastic-coupling contributions (2.236) and (2.237) in the
vertical and horizontal directions, respectively, can cause anisotropy of the shape of
microseismic clouds. In the case of fluid injection, the clouds of induced seismic
events should have a tendency to grow faster (i.e. to get longer) along the maxi-
mum principal stress direction. Schoenball et al. (2010) and Altmann et al. (2010)
observed this behavior on numerically simulated clouds of microseismic events.
Depending on the parameter n s the effects of poroelastic coupling can be strong
or weak. One of the main aims of the next section is to better understand the nature
of this parameter, and to estimate a realistic range of its values.
2.9 Several non-linear effects of poroelastic deformations

A non-linear effect of poroelastic deformations relevant for understanding the fluid-
induced seismicity is a stress and pore-pressure dependence of rock properties.
Especially interesting for applications are changes of elastic parameters and of
permeability as functions of applied loads. Understanding stress dependence of

elastic properties of rocks (including seismic velocities) is important, for example,
for interpretation of reflection seismic data, for analysis of 4D-seismic monitoring
of hydrocarbon reservoirs and for overpressure prediction based on seismic data.
Practical experience collected in exploration seismology shows that elastic non-
linearity of rocks is usually rather weak. This is purely a consequence of the
very small amplitude of seismic waves at the depth of typical hydrocarbon and
geothermal reservoirs (and hence, is a result of geometrical spreading). Seismic-
wave-propagation phenomena and other small deformation (incremental) effects
can be considered to be linear to a good approximation. On the other hand,
laboratory experiments on drained and fluid-saturated rock samples, as well as
seismic imaging and interpretation in depleting reservoirs and their overburden,
show significant changes in linear elastic parameters induced by variations in
confining stress and/or pore pressure.
Usually, in well-consolidated rocks, elastic strains due to moderate stress
changes (less than or of the order of 102 MPa) are on the order of 10−2 to 10−3 or
less. This is also the order of rotations and size changes in corresponding rock sam-
ples. Geometric non-linearity effects are even smaller; these are second-order terms
in respect to displacement gradients, and thus they are of the order of 10−4 –10−6 .
However, relative variations in elastic velocities can realistically reach the order
of 10−1 in typical laboratory experiments. On the other hand, elastic properties of
minerals (i.e. elastic moduli of the grain material) depend only weakly on stresses
in the stress range up to 108 Pa.
Natural cracks, fractures and grain-contact domains are most sensitive to
deformations. They are the most compliant parts of the pore space. Therefore, the
stress dependence of elastic moduli of drained rocks must be mainly controlled
by deformations of the pore space. This is the main reason of the physical (elas-
tic) non-linearity of rocks. Many authors related changes of contact geometries
(including spherical contact models) and crack geometries (including crack contact
models) with changes of the elastic properties; see, for example, Mindlin (1949),
Duffy and Mindlin (1957), Merkel et al. (2001), Gangi and Carlson (1996), and
others. We have seen in the previous sections that the theory of poroelasticity pro-
vides a general approach for describing deformations of the pore space. Let us
consider this effect in more detail.
2.9.1 Deformation of the pore and fracture space

Let us assume that in an initial stress state the internal surface p of the rock
sample was deformed. The geometry of the pore space can be characterized by
an initial location of the surface p plus its displacement due to the load. The
corresponding strain of the pore space averaged over the volume of the sample is
given by the quantity ξi j as defined in equation (2.22). If the rock is statistically

homogeneous then the quantity ξi j is independent of the size of the sample. Thus,
it is reasonable to assume that this quantity (along with the initial configuration of
p and the grain parameters) will control the elastic moduli of the drained rock.
Let us consider changes of the average strain ξi j due to changes of the load state
(δσiej ; δ Pp ). These changes of the strain will describe average changes of geometry
of the pore space. Changes of quantity ξi j satisfy the following rule:
ζi j δζi j δV
δξi j ≡ δ(− )=− − ξi j . (2.238)
V V V
The quantity ζi j is defined by equation (2.5). The quantity ξii is closely related to
the porosity. Indeed, for ξii equation (2.238) yields:
ζii δζii δV
δξii ≡ δ(− )=− − ξii . (2.239)
V V V
On the other hand, if V p0 is the volume of the pore space at state with ζi j = 0, then
the complete change of the porosity due to a change in the applied load will be
Vp V p0 − ζii δζii V p0 − ζii δV δζii δV
δφ = δ( ) = δ( )=− − =− −φ . (2.240)
V V V V V V V
Thus, the computing rule for a change of φ nearly coincides with the rule for a
change of ξii .
Let us consider Hooke’s laws permitting us to write these rules in forms of
stress–strain relations. We us assume that the grain material is homogeneous (in
other words we assume the validity of equation (2.11); a more-general considera-
tion can be found in Shapiro and Kaselow, 2005). Then, we obtain from equation
(2.21):
δζi j gr gr gr
− = (Sklij
dr
− Sklij )δσkl + (Sijkk
dr
− Sijkk − φ Sijkk )δ Pp . (2.241)
V
From equation (2.20) we obtain
δV
= Sklmm
dr
δσkl + (C dr − C gr )δ Pp . (2.242)
V
Substituting these two equations into equations (2.238) and (2.240) yields
gr gr
δξi j = (Sklij
dr
− Sklij − ξi j Sklmm
dr
)δσkle − (φ Sijmm − ξi j C gr )δ Pp , (2.243)
and, for the porosity,
gr
δφ = (Sklmm
dr
− Sklmm − φ Sklmm
dr
)δσkle . (2.244)
We observe that load-caused changes of the porosity depend on a single combi-
nation of the confining stress and pore pressure. This combination is the effective
stress, σkle . Again, this is a consequence of the condition C gr = C p . This condition

follows from equation (2.11) and is a consequence of the assumption of statistical
homogeneity of the grain material. It is also consistent with the Gassmann equation
(see also Brown and Korringa, 1975). In relation to this fact it is commonly said
that the effective-stress coefficient for the porosity (i.e. the coefficient in front of
the pore pressure in the effective stress) is equal to 1.
Finally, under a completely hydrostatic load (i.e. σkle = −δkl Pd , where Pd is
called the differential pressure), these equations are reduced to
gr gr
δξi j = −(Sllij
dr
− Sllij − ξi j C dr )δ Pd − (φ Sijkk − ξi j C gr )δ Pp , (2.245)
δφ = −(C dr − C gr − φC dr )δ Pd . (2.246)
Alternative derivations of equation (2.246) were presented by different authors
(see, for example, Zimmerman et al., 1986, Detournay and Cheng, 1993, and liter-
ature cited therein; see also Goulty, 1998, and Gurevich, 2004). This equation can
be also understood as a differential equation for the porosity as a function of the
differential pressure:
dφ
= C gr − (1 − φ)C dr . (2.247)
d Pd
Since the solid grain materials (common minerals) are nearly linear elastic
media, their compliances S gr and the compressibility C gr are practically indepen-
dent of loads. Thus, in equation (2.247), only two quantities are significantly stress
dependent: φ and C dr . In order to quantify the stress dependence of these quanti-
ties, at least one more equation relating them is required. This equation cannot be
formulated exactly because C dr is defined by the complete geometry of the pore
space rather than by φ only. Thus, further analysis requires involving empirical
observations and heuristic assumptions. Below we concentrate our consideration
on the stress dependence of the porosity of an isotropic rock in the case of a hydro-
static load (i.e. a change of a pore pressure and/or of a confining pressure) and
follow mainly the derivation proposed in Shapiro (2003). Analogous consideration
for anisotropic rocks and non-hydrostatic loads must involve a treatment of strain
ξi j . More details can be found in Shapiro and Kaselow (2005).
2.9.2 Stiff and compliant porosities

We separate total porosity φ into two parts
φ = φc + [φs0 + φs ] , (2.248)
where the first part, φc , is a compliant porosity supported by thin cracks and grain-
contact vicinities. According to laboratory observations, the compliant porosity
usually closes up by the differential pressure of a few tens of MPa. This corre-
sponds to the porosity with aspect ratio γ (a relationship between the minimal
and maximal dimensions of a pore) less than 0.01 (see Zimmerman et al., 1986).
The second part, [φs0 + φs ], is a stiff porosity supported by more-or-less iso-
meric pores (i.e. equidimensional or equant pores; see also Hudson et al., 2001,
and Thomsen, 1995) and worm-like pores (having their stiff cross-section of an
isomeric shape). The aspect ratio of such pores (or of their stiff cross-section,
respectively) is typically larger than 0.1. Such a subdivision of the porosity into
a compliant and stiff parts is frequently done in seismic rock physics. For exam-
ple, it is very similar to the definitions of the stiff and soft porosity of Mavko and
Jizba (1991).
In turn, we separate the stiff porosity into a part φs0 that is equal to the stiff
porosity in the case of Pd = 0, and to a part φs that is a change of the stiff porosity
due to a deviation of the differential pressure from zero. We assume that the relative
changes of the stiff porosity, φs /φs0 , are small. In contrast, the relative changes of
the compliant porosity (φc − φc0 )/φc0 can be very large, i.e. of the order of 1 (φc0
denotes the compliant porosity in the case of Pd = 0). However, both φc and φc0
are usually very small quantities. As a rule, they are much smaller than φs0 . For
example, in porous sandstones typical orders of magnitude of these quantities are
φs0 = 0.1, |φs | = 0.01 and φc can be less than, or of the order of, 0.01.
Under such circumstances, it is logical to assume the first, linear approximations
of the compressibility C dr as a function of the both parts of the porosity:
C dr (φs0 + φs , φc ) = C drs [1 + θs φs + θc φc ] , (2.249)
where C dr s is the drained compressibility of a hypothetical state of the rock with

a closed compliant porosity (i.e. φc = 0) and the stiff porosity equal to φs0 . The
acronym “drs” stands for “drained and stiff.” Becker et al. (2007) proposed to call
such a state of a rock a “swiss cheese model.” Furthermore,
1 ∂C dr 1 ∂C dr
θs = , θc = , (2.250)
C drs ∂φs C drs ∂φc
where the derivatives are taken at φs = 0 and φc = 0.

Approximation (2.249) implies that both quantities θs φs and θc φc are signifi-
cantly smaller than 1. Numerous laboratory experiments and practical experience
show that the drained compressibility depends strongly on changes in the compli-
ant porosity, and it depends much more weakly on changes in the stiff porosity. We
will express this empirical observation by the assumption
θs φs θc φc 1. (2.251)
The quantity θs can be approximately estimated by using the upper bound of

Hashin and Strikman (1963) for K dr . Such an estimate corresponds to a small
concentration of spherical voids in a solid matrix (see also Mavko et al., 1998):
3 K gr
θs ≈ 1 + . (2.252)
4 μgr
This shows that θs has values of the order of 1. Thus, θs φs is a quantity of the order
of 0.01.
An idea about the order of magnitude of θc can be obtained from results of
various effective-medium theories for penny-shaped cracks, which usually coin-
cide in the limit of φc −→ 0 (see Mavko et al., 1998, for effective bulk-modulus
formulations in self-consistent or differential effective-medium approximations):
K gr (3K gr + 4μgr )
θc ≈ . (2.253)
πγ μgr (3K gr + μgr )
This estimate shows that θc is usually of the order of 1/γ . Therefore, θc can be of
the order of 102 or larger. Thus, in spite of a very small porosity φc the quantity θc φc
can reach the order of 0.1 or become even larger. If so, the approximation (2.249)
is further simplified as follows:
C dr (φs0 + φs , φc ) = C drs [1 + θc φc ] . (2.254)
Using this approximation, and neglecting φ in comparison with 1, we obtain the
following relationship instead of equation (2.247):
dφs dφc
+ = C gr − C drs − θc φc C drs . (2.255)
d Pd d Pd
We further assume that stiff-porosity changes with stress are independent of the
changes of the compliant porosity. This also means that changes of the stiff porosity
are independent of whether the compliant porosity is closed or not. If the compliant
porosity is closed then φc = 0 and we obtain from (2.255)
dφs
= C gr − C drs . (2.256)
d Pd
Furthermore, if the assumption above is valid then this relationship will be valid
also for an arbitrary (however, because of other assumptions, small) φc . Therefore,
dφc
= −θc φc C drs . (2.257)
d Pd
Solution of equations (2.256) and (2.257) provides us with the following approxi-
mations of the stress dependence of the stiff and compliant porosities:
φs = Pd (C gr − C drs ), (2.258)
φc = φc0 exp (−θc Pd C drs ). (2.259)

Note that equation (2.258) is not valid for very large Pd . This is due to the fact
that in equation (2.254) we neglected the stiff-porosity dependence of the com-
pressibility C drs , which becomes equal to C gr if Pd −→ ∞. The validity of
such a simplification and the validity of equation (2.258) are restricted by the
condition (2.251). For very high stresses, the stiff porosity will also obey an
exponentially saturating decreasing behavior.
2.9.3 Stress dependence of elastic properties

Usually, the dependence of a seismic velocity on a load is phenomenologically
described by the following simple relationship (Zimmerman et al., 1986; Eberhart-
Phillips et al., 1989; Freund, 1992; Jones, 1995; Prasad and Manghnani, 1997;
Khaksar et al., 1999; Carcione and Tinivella, 2001; Kirstetter and MacBeth, 2001):
c(Pd ) = Ac + K c Pd − Bc exp (−Pd Dc ), (2.260)
The coefficients Ac , K c , Bc and Dc of equation (2.260) are fitting parameters for a
given set of measurements. It is often observed that equation (2.260) provides very
good approximations for velocities in dry (see, for example, Figure 2.9) as well as
in saturated rocks. Below we will derive this equation and clarify the nature of the
fitting parameters.
Substituting equations (2.258) and (2.259) into equation (2.249), we obtain

C dr (Pd ) = C drs 1 − θs (C drs − C gr )Pd + θc φc0 exp (−θc Pd C drs ) . (2.261)
For the bulk modulus this gives approximately

K dr (Pd ) = K dr s 1 + θs (C drs − C gr )Pd − θc φc0 exp (−θc Pd C drs ) . (2.262)
Using for the skeleton shear modulus μdr , an expansion similar to (2.249) yields

μdr (Pd ) = μdr s 1 + θsμ (C drs − C gr )Pd − θcμ φc0 exp (−θc Pd C drs ) , (2.263)
where μdr s is the shear modulus of the dry rock in the case of a closed com-
pliant porosity and stiff porosity equal to φs0 (i.e. the case of a “swiss cheese
model”) and
1 ∂μdr 1 ∂μdr
θsμ = − , θcμ = − . (2.264)
μdr s ∂φs μdr s ∂φc
By analogy with the quantity θs the quantity θsμ can be estimated using the upper
Hashin–Strikman bound of μdr :
6K gr + 2μgr
θsμ ≈ 1 + . (2.265)
9K gr + 8μgr
403Cli515:P-wavesD = 0.0264 [1/MPa]
6.8
V11 best−fit
V11 emp.
6.6 V22 best−fit
V22 emp.
6.4 V33 best−fit
V33 emp.
6.2
V [km/s]
6
5.8
5.6
5.4
5.2
5
0 100 200 300 400 500 600
P [MPa]
403Cli515:S-wavesD = 0.0264 [1/MPa]
4
V12 best−fit
V12 emp
V13 best−fit
3.8 V13 emp
V21 best−fit
V21 emp.
V23 best−fit
3.6
V23 emp
V31 best−fit
V [km/s]
V31 emp.
3.4 V32 best−fit
V32 emp
3.2
2.8
0 100 200 300 400 500 600
P [MPa]
Figure 2.9 Measurements (dots) of quasi P-wave velocities (top) and quasi S-
wave velocities (bottom) of a dry, metamorphic rock sample from the German
KTB deep borehole. Velocities were measured in three orthogonal directions in
laboratory experiments by Kern and Schmidt (1990) and Kern et al. (1994). Best-
fit lines corresponding to equation (2.260) with different coefficients Ac , K c , Bc
reveal nearly the same value of the parameter, Dc = 0.026 MPa−1 . (After Shapiro
and Kaselow, 2005.)
We see that the quantity θsμ has the order of magnitude of 1. Again, by using
effective-media theories for small concentrations of penny-shaped cracks, we can
estimate the order of magnitude of θcμ (see Mavko et al., 1998, for effective
shear modulus formulations in self-consistent or differential effective-medium
approximations):

1 4(3K gr + 4μgr )(9K gr + 4μgr )
θcμ ≈ 1+ . (2.266)
5 3πγ (3K gr + μgr )(3K gr + 2μgr )
This estimate shows that for small γ , the quantity θcμ is usually of the order of 1/γ .
Approximations similar to (2.262) and (2.263) can be obtained for the

undrained bulk modulus and for the velocities of P- and S-waves (for details, see
Shapiro, 2003). A comparison of results for elastic moduli (2.261)–(2.263) with
equation (2.260) shows that they all have the same functional form. This also leads
to the same form of the stress dependencies of velocities. For example, for the
P-wave and S-wave velocities in drained rocks, c Pdr and c Sdr , the corresponding
relations will have the following forms:
c Pdr (Pd ) = A P + K P Pd − B P exp (−Pd D P ), (2.267)
c Sdr (Pd ) = A S + K S Pd − BS exp (−Pd D S ). (2.268)

Using results (2.265) and (2.266) we can estimate the order of terms in equa-
tions (2.262) and (2.263). If the differential pressure is smaller than, or of the
order of, 102 MPa and Cdr s is of the order of 10−5−10−4 MPa−1 , the terms
θsμ (C drs − C dr )Pd and θs (C drs − C dr )Pd in equations (2.262) and (2.263) will
be of the order of 0.001−0.01. These terms can often be neglected because the
other two terms in equations (2.262) and (2.263) are of the order of 0.1 and 1.
Therefore, if Pd is measured in MPa, both terms with coefficients K P and K S in
equations (2.267) and (2.268) are of the order of 1 and, thus, can be neglected in
comparison with other, much larger terms. This fact is in agreement with obser-
vations of Zimmerman et al., (1986), Khaksar et al., (1999), and Kirstetter and
MacBeth (2001).
The modified equations have similar forms for the both, P- and S-wave
velocities:
c Pdr (Pd ) = A P − B P exp (−Pd D P ), (2.269)
c Sdr (Pd ) = A S − BS exp (−Pd D S ). (2.270)

These equations also keep their forms for saturated rocks. Moreover, we
should expect that, for dry as well as for saturated isotropic rocks, in the first
approximation all coefficients Dc are identical:
D P = D S = θc C drs . (2.271)
This is also in good agreement with laboratory estimates of these quantities for
porous sandstones (see Figure 2.10). Moreover, an approximate universality of
coefficients Dc has been observed for different velocities in anisotropic rocks
(see Ciz and Shapiro (2009) and Figure 2.9) and even for elastic stiffnesses and
electrical resistivity (Kaselow and Shapiro, 2004).
Summarizing, to the first approximation the seismic velocities as well as
the porosity depend on the differential pressure, i.e. the difference between the
confining pressure and the pore pressure. The stress dependence of the geometry of
0.1
DS(1/MPa)
0.01
0.01 0.1
DP(1/MPa)
Figure 2.10 Coefficient D S versus D P for several sandstones (saturated as well as

dry). The straight line is given by (2.271). Data have been collected from different
literature sources: Eberhart-Phillips et al. (1989); Freund (1992); Jones (1995);
Khaksar et al. (1999). (After Shapiro, 2003.)
the pore space controls the elastic moduli and velocity changes with stress. Here,
the compliant porosity, which can be just a very small part of the total porosity,
plays the most important role. Closing compliant porosity with increasing dif-
ferential pressure explains the experimentally observed exponentially saturating
increase of seismic velocities. Coefficients of this relationship are defined by the
compliant-porosity dependence of the drained bulk modulus. The coefficients of
equations (2.267) and (2.268) can be used to compute such characteristics of rocks
as φc0 , θcμ and θc .
The dimensionless quantity θc defines the sensitivity of the elastic moduli to
the differential pressure. Shapiro (2003) proposed calling it the elastic piezo-
sensitivity. The piezo-sensitivity is a physical property of a rock. From the
derivation above it is clear that it is controlled by the compliant porosity. More-
over, it is approximately proportional to an averaged (effective) reciprocal aspect
ratio of the compliant porosity. Taking into account the data of Figure 2.10 we
obtain that the realistic range of the magnitude orders of the piezo-sensitivity for
sandstones is 102 –104 . For low-permeable tight rocks like the crystalline rocks of
the German KTB borehole, basalts, granites and shale (including organic reach
shale), the piezo-sensitivity is of the order of 102 or even smaller. This can be
estimated from values of Dc obtained in different works (Ciz and Shapiro, 2009;
Kaselow et al., 2006; Becker et al., 2007; Pervukhina et al., 2010) and also from
frequently reported observation that elastic properties of some shales are nearly
stress independent in the range of loads below 30 MPa.
2.9.4 Non-linear nature of the Biot–Willis coefficient α

The piezo-sensitivity can be related to the poroelastic coefficient α. Indeed,
substituting (2.254) into (2.35) we obtain:
(1 − θc φc )
α =1− . (2.272)
C drs K gr
We can obtain upper and lower bounds for α. Following Zimmerman et al. (1986)
and Norris (1989), we use the Hashin–Strikman bounds for the drained compress-
ibility C drs . Its upper bound is infinitely large and yields α = 1. Its lower bound
corresponds to spherical void inclusions distributed in the grain material so that
the porosity of the resulting composite is φs0 . As a result we obtain the following
bounds for α:
(1 − φs0 ) 2θc φc + 3φs0
1 ≥ α ≥ 1 − (1 − θc φc ) ≥ ≈ 1.5φs0 + θc φc ,
1 + 4 φs0 K gr /μgr
3 2 + φs0
(2.273)
where we assumed the positiveness of Poisson’s ratio for the rock-grain material
(i.e. K gr /μgr ≥ 2/3) and have neglected terms with products of porosities. If
φc = 0 we obtain
3φs0
1≥α≥ ≈ 1.5φs0 . (2.274)
2 + φs0
These are the bounds of α derived by Zimmerman et al. (1986). However, the val-
ues of α usually observed (see, for example, table 4 of Detournay and Cheng, 1993)
are much larger than just 1.5φ or even 2φ (the latter corresponds to situations with
the grain material Poisson’s ratio close to 0.25). Equation (2.273) clearly explains
the reason of such observations. The compliant porosity increases the lower bound
of α. The greater is the piezo-sensitivity and/or compliant porosity of rocks the
greater is the lower bound and, correspondingly, the greater the quantity α tends
to be. Obviously, the smaller is the effective aspect ratio of the compliant poros-
ity the higher is the lower bound of α. The upper bound of α is always 1, of
course.
On the other hand, the piezo-sensitivity is related to the non-linear elastic moduli
of rocks. Indeed, for small values of Pd we obtain from equation (2.262)
K dr (Pd ) = [1 + β K Pd ]/C drs , (2.275)
where the coefficient

β K = θc2 φc0 C drs (2.276)
is equal to a rational function of the bulk modulus and of the non-linear elastic
moduli (see Zarembo and Krasilnikov, 1966, pp. 299–309). It is clear that the
larger the piezo-sensitivity and the compliant porosity, the larger the elastic non-
linearity of rocks. Taking into account the values of the piezo-sensitivity following
from Figure 2.10, we can expect that for well-consolidated sedimentary rocks the
non-linearity is 102 –104 times larger than for such materials as metals, where the
coefficient β K is roughly of the order of 1/K .
2.9.5 Magnitude of the poroelastic-stress coupling

Equations (2.276), (2.272) and (2.273) show that coefficient α is closely related
to the piezo-sensitivity and to the coefficient β K and, thus, to the non-linear elas-
tic moduli of drained rock. This fact can be used to estimate the strength of the
poroelastic-stress coupling. Indeed, in Sections 2.6–2.8 we have seen that the
stress effects of the poroelastic coupling are controlled by parameter n s , which
is propotional to the Biot–Willis coefficient (see also (2.167)):
n s = αn , (2.277)
where we used notation

1 − 2ν dr
n ≡ . (2.278)
2(1 − ν dr )
Figure 2.11 shows n as a function of the drained Poisson’s ratio. We conclude
that generally for realistic rocks with positive Poisson’s ratio, n s < 0.5.
In the following chapters of this book we will be often interested in low-
permeability crystalline and sedimentary tight rocks, where stimulation of the fluid
mobility is necessary. The stiff porosity of such rocks belongs effectively to the
solid material composing their drained skeleton. Thus, the product K gr Cdr s is
approximately equal to 1. Then from (2.272) and (2.259) we obtain:
α ≈ θc φc = θc φc0 ex p(−θc Cdr s Pd ). (2.279)
Generally, 1 θc φc0 . The product θc φc0 is equal to the largest correction of the
bulk compressibility of drained rock due to the presence of compliant porosity.
For low-permeability tight rocks this correction is usually lower than 0.3. For such
rocks Poisson’s ratio is rather close to 0.25 (see Figure 2.11, top) and, therefore,
usually n s < 0.1.
Moreover, equation (2.279) implies that α decreases with increasing depth.
Usually for low-permeability tight rocks one can expect that φc0 is of the order
of 0.001 or smaller. Even for relatively permeable rocks like sandstones the com-
pliant porosity can also be of the order of 0.001 (see, for example, Pervukhina
et al., 2010). Realistic values of θc for tight rocks are of the order of 10–1000 (see,
for example, estimates obtained for different rocks including shales: Pervukhina
0.5
0.4
0.3
0.2
0.1
0
0 0.1 0.2 0.3 0.4 0.5
dr
v
0.20
0.18
0.16
0.14
0.12
0.10
0.08
0.06
0.04
0.02
10 50 100 500 1000

θc
Figure 2.11 Poroelastic-stress coefficient and its components. Top: the function
n ≡ (0.5−ν dr )/(1−ν dr ) of the drained Poisson’s ratio. Bottom: the poroelastic-
stress coefficient n s as a function of the piezo-sensitivity, θc . The parameters of
the curves are: n = 0.3, φc0 = 0.001 and K dr s = 30 GPa, and Pd is equal to 10,
20, 30, 40 and 50 MPa from the upper curve to the lower curve, respectively. In
the case of hydrostatic conditions these differential pressures roughly correspond
to depths of 0.5, 1, 1.5, 2 and 2.5 km, respectively.
et al., 2010; Ciz and Shapiro, 2009; Kaselow et al., 2006). Figure 2.11 (bottom)
shows some estimates of the poroelastic-stress coefficient. For a realistically cho-
sen Cdr s and for the depth greater than 1.5 km, the poroelastic-stress coefficient
is significantly smaller than 0.1 (nearly the maximum value). It is quickly vanish-
ing with depth. Correspondingly, the role of the poroelastic-stress coupling will
quickly vanish with depth.
In the previous sections we have seen that, in the case of fluid injection, a
strong poroelastic coupling will lead to stronger growth of a cloud of fluid-induced
seismicity along a direction of the maximum tectonic stress (maximum absolute
value). In the case of fluid extraction this will lead to a possible rock failure,
firstly along the minimum tectonic stress (minimum absolute value). However,
the magnitudes of stress perturbations caused by poroelastic coupling are con-
trolled by the poroelastic-stress coefficient. Our analysis in this section shows that,
for low-permeability tight rocks at depths of 1.5 km and more, this coefficient is
smaller than 0.1. Under such conditions the pore-pressure perturbation will make
a dominant contribution to poroelastic modifications of a failure-criterion stress.
Thus, in the following we will usually neglect the poroelastic coupling effects (see
also Section 3.4).
2.9.6 Effective-stress coefficients

Confining stress and pore pressure are applied to different surfaces of a porous
sample. Their impacts on the rock can be to some extent mutually compensating or
enhancing. In the previous sections we have seen that different quantities charac-
terizing poroelastic media depend on the applied confining stress and on the pore
pressure in different manners. Generally this dependence can be expressed as a
function of the following combination of the loads:
q q
L i j ≡ −σi j − ρi j Pp . (2.280)
q
Berryman (1992, 1993) considered different forms of load combinations L i j as
q
effective-stress rules for the corresponding property q of a rock. The quantity ρi j
is called an effective-stress coefficient for the physical quantity q. It is clear that,
generally, this coefficient can be a tensor. For example, for the strain tensor kl of
a porous fluid-saturated rock sample (i.e. a strain of the skeleton) equation (2.31)
q
gives ρi j = αi j .
Further, we will follow Berryman (1992, 1993) and consider a simplified sit-
uation of a hydrostatic load defined by a confining pressure Pc and by a pore
pressure Pp . We will assume also a statistically homogeneous grain material. We
will also consider load-dependent quantities such as scalars. In this case, effective
stress (2.280) for a given physical quantity q reduces to the following quantity:
L q = Pc − ρq Pp . (2.281)
For example, equation (2.37) shows that for the sample dilatation (i.e. skeleton
dilatation) the effective-stress coefficient is equal to ρ = α. Equation (2.43) shows
that, for the quantity χ , which is a change of the pore-space volume in a unit rock
volume caused by a fluid-mass migration, the effective-stress coefficient is equal to
ρχ = 1/B, where B is the Skempton coefficient. Finally, equation (2.246) shows
that the effective-stress coefficient for the connected porosity, ρφ is 1. Moreover,

equations (2.241)–(2.246) show that, in the case of small porosity and small strains,
the effective-stress coefficient for the geometry of the pore (and fracture) space is
close to 1.
The effective-stress coefficient for the rock failure is not directly related to
poroelastic coefficients ρq . However, the failure of a rock must be related to defor-
mations of its pore (and fracture) space. Thus, it must be related to ρq describing
load dependence of the pore-space geometry. Terzaghi (1936) concluded from
heuristic reasoning that the effective-stress coefficient for a load combination con-
trolling the failure of saturated rocks (we will denote this combination as L f ) is
f
equal to 1. This corresponds to ρi j = δi j in equation (2.280) and to ρ f = 1 in
equation (2.281). Experimental observations and the fact that the stress concentra-
tions at microcracks are proportional to the differential pressure also support the
assumtion that ρ f = 1 (see Jaeger et al., 2007).
2.9.7 Stress dependence of permeability

Deformations of the pore space are also related to the stress dependence of the
permeability. To understand the effective-stress law for the permeability, Berry-
man (1992) proposed the following argument. The physical dimension of the
permeability is (length)2 . Thus, uniform rescaling of the size of an isotropic porous
sample (leading also to the corresponding rescaling of sizes of pores) increases or
decreases permeability proportionally to (volume)2/3 . Note that such a geometric
transformation of the pore sample does not change its porosity. In other words, this
rescaling does not change specific geometric relations between the solid and the
pore space. However, additionally to the sample scale, permeability must depend
on a non-dimensional measure of mutual geometric relations between the solid
and the pore space (i.e. the relative positioning of grains). A power-law function
of the porosity frequently seems to be such an adequate measure. For example, it
is in agreement with a Kozeny–Carman-type of relation between the porosity and
permeability for laminar flow in tube-like pores and other pore channels of simple
shapes (see Coussy, 2004). Taking this into account we obtain
k ∝ φ n V 2/3 , (2.282)
where n is a nearly empiric exponent close to 4 or larger.
Equation (2.282) implies that the permeability is not just a simple function of
the differential pressure. This is clear from the fact that the permeability depends
not only on the porosity but also on a characteristic length. In equation (2.282), the
characteristic length is related to the sample volume. On the other hand, our discus-
sion above indicates that the effective-stress law for the volume includes α as the
effective-stress coefficient. Therefore, the permeability depends on a combination

of the load components Pp and Pc that is more complex than the differential
pressure.
Using relation (2.282), Berryman (1992) shows that the effective-stress coeffi-
cient for permeability ρk must be smaller than 1 and it is a function of α, φ and n.
However, real rocks can show ρk lower, equal to or larger than 1 (Berryman, 1992;
Al-Wardy and Zimmerman, 2004; Li et al., 2009). The reason for this is the fact
that equation (2.282) has been derived from too-general dimensionality considera-
tions. By such considerations the main dimensional parameters are the volume of
the sample and the volume of its pore space. Thus, the length scale in (2.282) can
be substituted by a pore diameter or a grain diameter instead of a scale derived
from the macroscopic volume. However, two length scales (the pore diameter and
the sample size) are not sufficient to describe the pressure-dependent permeability.
In other words, equation (2.282) treats the porosity in a too-general way. It does not
specify which porosity component mainly controls the permeability. The total con-
nected porosity can be substituted in this equation by any other porosity function.
Later we will be especially interested in situations where the main changing
load component is an increasing pore pressure (i.e. a decreasing differential pres-
sure) leading to an expansion of the pore space. Equation (2.282) can be taken
as a basis for modeling pressure-dependent permeability if we use additional
information on the pore-space geometry.
For example, in the previous sections we argued that the stress dependence
of drained elastic properties is mainly controlled by the compliant porosity. The
stiff porosity is frequently insignificant. Roles of compliant and stiff parts of the
pore space can also be very different in the pressure dependence of permeability.
To model such a pressure dependence we combine equation (2.282) with equa-
tions (2.248), (2.259) and (2.258). To account for possibly different contributions
of the stiff and compliant porosities in the permeability we take these porosity
components with different weighting factors (s and c , respectively) introduced
ad hoc:
n
k ∝ s (φs0 − Pd (C drs − C gr )) + c φc0 exp (−θc Pd C drs ) V 2/3 . (2.283)
Note that it is sufficient to work with a single relative weighting factor only, e.g.
c /s . However, we keep both factors for convenience.
From relation (2.283) we can derive several simple models. By moderate loads
the volume V will not change a lot. Thus, we can include it approximately into a
proportionality factor. Thus, we do approximate the effective stress by the differen-
tial pressure. Let us further assume a rock where the permeability is controlled by a
stiff porosity only (i.e. c = 0). For example, this can be the case for high-porosity
sand, where compliant pores are mainly confined to grain contacts and do not sig-
nificantly contribute to permeability (which is controlled by stiff pore throats). For
c = 0 we obtain
C drs − C gr
k ∝ (1 − Pd )n . (2.284)
φs0
This function resembles theoretical models derived by Gangi (1978) and
Walsh (1981).
Let us further assume that the permeability is controlled by stiff porosity, but
the initial (connected) stiff porosity is vanishingly small (i.e. φs0 = 0). Then the
increasing pore pressure will open new connected stiff pores. Under such condi-
tions Pd < 0. Equation (2.284) is still applicable. We obtain a power-law model of
the permeability:
k ∝ (−Pd )n . (2.285)
Conversely, assume that the contribution of the stiff porosity in the connected
porosity is vanishingly small. This will lead to s = 0. Then the permeability
is controlled by the compliant porosity. This corresponds, for instance, to fractured
carbonates, where compliant voids (fractures) provide the only conduits between
isolated vugs. This yields
k ∝ exp(−κ Pd ). (2.286)
Here κ is the so-called permeability compliance. Equation (2.283) implies that
κ = θc C drs n. (2.287)
Both types of pressure dependence (2.285) and (2.286) can be considered as
simple modeling alternatives. Then quantities n and b p are fitting constants. For
example, function (2.286) has been used in experimental studies (Li et al., 2009).
2.10 Appendix. Reciprocity-based relationship between compliances

of porous media
In this appendix we will derive equation (2.13). We apply the reciprocity theorem
(Amenzade, 1976; Brown and Korringa, 1975):
I = [δu i (x̂)δ τi (x̂) − δ u i (x̂)δτi (x̂)]d 2 x̂ = 0, (2.288)

g
where the total surface of the linear grain material is introduced: g = c + p .

Note that we assume the existence of a vanishing thin film sealing pores on the
external surface of the rock specimen. Any influence of this film on the drained
compressibilities is neglected. Such an enveloping thin film is assumed to allow an
2.10 Appendix. Reciprocity-based relationship 117
application of independent forces to c and p (i.e. a jacketed sample as shown in

Figure 2.2) Further, δτi and δ τi are components of two systems of traction incre-
ments acting on the surface g in two different hypothetical experiments. These
small traction increments cause small displacements of the surface g . Quantities
δu i and δ u i are components of these displacements, respectively.
Let us consider the following pair of traction increments uniformly distributed
on c or p :
τi = δσi j n j on c and τi = 0 on p , (2.289)
and
τi = 0 on c τi = δσi j n j on p .
f
and (2.290)
The reciprocity theorem then gives:
δu k (x̂)n l d 2 x̂.
f
δσi j δu i (x̂)n j d 2 x̂ = δσkl (2.291)
p c
Assuming that the traction increments and displacements are small, and using the
notations from equations (2.3) and (2.5), equation (2.291) can be written as
f
δσi j δζi j = δσkl δηkl . (2.292)
f
Since this holds for any δσi j and δσkl (i.e. we can select several components being
non-vanishing only) we have:

∂ζi j ∂ηkl
= . (2.293)
∂σkl σ f f
∂σi j σ
f f
By changing independent variables from σi j
and σi j to σi j and σiej (see defini-
tion (2.6)), we obtain

∂ζi j ∂ζi j ∂ηkl ∂ηkl
= = − . (2.294)
∂σkl σ f ∂σkle σ f f
∂σi j e ∂σiej f
σ σ
Dividing this by the volume, V , we obtain (2.13).

3
Seismicity and linear diffusion of pore pressure
In this chapter we consider situations of fluid injections with the injection pressure
(i.e. the bottom hole pressure or the pressure at the perforation borehole interval)
smaller than the absolute value of the minimum principal compressive stress, σ3 .
Then, usually, permeability enhancement on the reservoir scale is significantly
smaller than one order of magnitude. In such a situation the behavior of the
seismicity triggering in space and in time is mainly controlled by a linear process
of relaxation of stress and pore pressure.
We will start this chapter by considering fluid-injection experiments in
crystalline rocks at the German Continental Deep Drilling site KTB (Konti-
nentale Tiefbohrung). Results of these experiments are helpful for identifying
factors controlling fluid-induced seismicity. Then we consider kinematic features
of induced seismicity such as its triggering front and its back front. They pro-
vide envelopes of clouds of microseismic events in the spatio-temporal domain.
We further consider features of the triggering front in a hydraulically anisotropic
medium.
We introduce several approaches for the quantitative interpretation of
fluid-induced seismicity. These approaches can be applied to characterize hydraulic
properties of rocks on the reservoir scale. Thus, they can be useful for construct-
ing models or constraints for reservoir simulations. These approaches are based
on the assumption of a linear fluid–rock interaction. Some of these approaches
have a heuristic nature. These are the triggering-front-based estimation of hydraulic
properties of rocks (in the next chapter we will observe that in some cases of a non-
linear pressure diffusion the triggering front can be introduced exactly) and the
related eikonal-equation-based approach for characterization of spatially heteroge-
neous hydraulic diffusivity. We will also consider a dynamic property of induced
seismicity: its spatial density. A related issue of statistical properties of the strength
of pre-existing defects will be a subject of our consideration too.
118
3.1 Case study: KTB

The German KTB site (see a series of papers with an introductory overview
of Emmermann and Lauterjung, 1997) is located in Windischeschenbach
(Southeastern Germany, Bavaria). It was placed near the western margin of the
Bohemian Massif at the contact zone of the Saxo-Thuringian and the Moldanubian
tectonic units. The corresponding crystalline crustal segment is mainly composed
of metabasites and gneisses. The site includes two boreholes, the pilot and the
main one, drilled between 1987 and 1994. The pilot hole reached a depth of 4 km,
whereas the main hole penetrated the crust down to 9.1 km.
Three large-scale series of borehole-fluid injection-extraction experiments were
performed at the site with the aim of investigating fluid-transport processes and
crustal stresses.
The first fluid-injection experiment was carried out in 1994 (Zoback and Harjes,
1997). The duration of the injection was approximately 24 h. About 200 m3 of
KBr/KCl brine were injected into a 70 m open section of the main borehole in
the depth range 9030–9100 m (with in situ temperature of 260 ◦ C). During the
active injection period the flow rates varied approximately from 50 to 550 l/min and
the corresponding well-head pressure varied approximately from 10 to 50 MPa.
The downhole injection pressure was approximately 130 MPa larger due to the
static pressure of the borehole-fluid column.
The seismic observation system included 73 surface stations and one three-
component borehole seismometer. Approximately 400 microearthquakes were
recorded. All events were considered to be induced by the injected fluids: 94
of these earthquakes could be localized with respect to master events with a
relative location accuracy of several tens of meters (Zoback and Harjes, 1997; Jost
et al., 1998). Later, the data were precisely relocalized by using various hypocenter
location improvements (Baisch et al., 2002). The seismically active zone com-
prised a volume of approximately 0.35 km3 around the bottom of the borehole.
By analyzing modeling results, Zoback and Harjes (1997) estimated that a small
increase in pore pressure was sufficient to trigger the earthquakes. A perturbation
of the pore pressure of 1 MPa corresponds roughly to the distance of 30 m from
the injection interval. However, there were many significantly more-distant events.
Surprisingly, only events above a depth of 9 km were observed. Hypothetical rea-
sons for this were proposed by Zoback and Harjes (1997). One of the reasons is
the possible proximity of the brittle–ductile transition zone at this depth and, as a
consequence, the presence of an impermeable lower half-space. On the other hand,
it is possible that, at greater depth, the level of the differential stress is significantly
smaller than the rocks’ frictional strength. Thus significantly larger pore-pressure
perturbations would be required for inducing earthquakes.
120 Seismicity and linear diffusion of pore pressure
The second fluid-injection experiment at the KTB was performed in the year
2000 (Baisch et al., 2002). This was a two-month-long injection. It was designed
to enable fluid migration and to perturb the pore pressure farther away from the
injection interval than in the previous experiment. About 4000 m3 of water were
injected at the well head of the main borehole at an injection rate of 30–70 l/min
and a well-head pressure between 20 and 30 MPa. It was assumed that the main
borehole was hydraulically isolated at least down to 6 km depth. However, due to
several leakages in the borehole casing, about 80% of the fluid was already injected
between 5.35 and 5.4 km depth.
A temporary seismic network consisting of a borehole seismometer in the pilot
hole at 3827 m depth and 39 surface stations was installed (Baisch et al., 2002).
Nearly 2800 microseismic events were detected, of which 237 were localized with
an accuracy better than 100 m on average. Seismic events concentrated at two
depth levels: 5.0–6.0 km (81% of total seismicity) and 8.8–9.2 km (11% of total
seismicity).
In June 2002 the third series of long-term fluid experiments (2002–2005) was
started (Kümpel. et al., 2006; Erzinger and Stober, 2005; Shapiro et al., 2006a).
Features of particular interest of the research program were two dominant fault
systems encountered at 7.2 km and 4.0 km depths. These two fault systems were
identified as seismic reflectors at the KTB site and denoted as SE1 and SE2 reflec-
tors, respectively (Hirschmann and Lapp, 1995; Simon et al., 1996; Harjes et al.,
1997; Buske, 1999). The first major experiment was a one-year-long fluid produc-
tion test in the KTB pilot hole (June 2002–June 2003). A total volume of 22 300 m3
of saline crustal fluids (with temperature of 119 ◦ C in situ) was produced from the
open hole section (3850–4000 m, approximately). The final draw down was 605 m
below the surface. The corresponding fluid yield was 58 l/min. The KTB main hole
was equipped with a seismometer installed at 4000 m depth and two water-level
sensors. The fluid level in the main hole – at 200 m distance from the pilot hole –
steadily fell from zero to 50 m below the surface, indicating a hydraulic connec-
tion between the pilot and the main hole. During the fluid-production phase of the
experiment, induced seismicity was absent. Hydraulic permeability was estimated
to be around 2 × 10−15 m2 (Stober and Bucher, 2005; Gräsle et al., 2006).
After 12 months of hydraulic recovery, a fluid injection test in the pilot hole was
started in June 2004. Over ten months, 84 600 m3 of water were injected into the
open hole section at 4 km depth of the pilot hole, where the SE2 reflector inter-
sected the borehole. The injection rate was 200 l/min on average, at about 10 MPa
well-head injection pressure. The fluid level in the main hole clearly responded
to the injection in the pilot hole. In October 2004, the main hole became artesian
and produced some 1 m3 of water per day. Significant induced seismicity started in
September 2004 and increased slowly (see Figure 3.1).
27.05.04 01.06.05
10 160
(a)
# Events (cumulative)
9 140
8
120
# Events/Day
7
6 100
5 Surface stations 80
4 60
3
40
2
1 20
0 0
0 25 50 75 100 125 150 175 200 225 250 275 300 325 350
140 (b) 2800
120 2400
# Events/Day
100 3485 m 2000

80 1600
60 Main-hole sonde 3500 m 1200
No data from
40 sonde 800
20 1950 m 400
0 0
0 25 50 75 100 125 150 175 200 225 250 275 300 325 350
S-P travel time at HBR [s]
(c)
0.3 0.3
0.2 0.2
0.1 0.1
0.0 0.0
0 25 50 75 100 125 150 175 200 225 250 275 300 325 350
100 000
Injection volume [m3]

140 (d) Volume
Pressure [bar]
80 000
120
Pressure 60 000
100
40 000
80
20 000
60
0
0 25 50 75 100 125 150 175 200 225 250 275 300 325 350
Days since start of injection experiment (27/05/04)
Figure 3.1 Microseismicity during the 2004–2005 KTB injection experiment.

(a) Events recorded by the surface stations. (b) Events recorded by the geophone
located in the main borehole. (c) Time differences between arrivals of S- and
P-waves. Changes in the locations of the borehole geophone are clearly seen
from the plot. (d) The injection pressure along with the cumulative volume of
the injected water. The time when the amount of previously extracted fluid was
re-injected is marked in part (a) of the figure by the inverse triangle on the top,
between the days 100 and 125. Another triangle (between the days 325 and 350)
marks the injection termination. (Modified after Shapiro et al., 2006a.) A black
and white version of this figure will appear in some formats. For the color version,
please refer to the plate section.
During the injection experiment, about 3000 microseismic events were detected
by the borehole sonde and about 140 events were localized by using the
borehole sonde and a specially deployed monitoring array of three-component
seismometers (five sensors in 25–50 m shallow boreholes and eight surface
stations).
One important observation of the 2002–2005 KTB hydraulic experiments was

the absence of seismicity during the phase of fluid extraction. Moreover, the
induced seismicity started during the injection phase approximately at the time
when the fluid volume previously extracted had been compensated by the injection.
It was approximately 110 days after the start of the injection (see Figure 3.1).
Horizontal projections of the located events compose a NW–SE elongated
zone. This zone is nearly parallel to the Franconian Lineament. Before and dur-
ing the drilling phase at the KTB site, intensive seismic studies were carried
out. From a 2D seismic survey (KTB8502), a sharp northeast-dipping seismic
reflector (SE1) was identified in seismic profiles (Simon et al., 1996; Harjes
et al., 1997; Buske, 1999). This reflector is regarded as the continuation of
the Franconian Lineament through the upper crust. A pre-stack Kirchhoff depth
migration of the KTB8502 profile as well as of a 3D seismic reflection sur-
vey (ISO89-3D) was presented by Buske (1999). During this survey, an area
of about 21 km × 21 km, with the main borehole located at the center, was
investigated. Figure 3.2 shows the relevant parts of this data set after migration.
Above the SE1 reflector and quasi-parallel to it, a slightly reflecting linear struc-
ture intersecting the boreholes approximately at the depth of 4 km can be also
seen. This is the SE2 reflector. The induced seismicity of the 2004–2005 injec-
tion experiment is directly related to this reflector (Figure 3.2). This is even
more evident in Figure 3.3, where a horizontal slice of the reflectivity at the
depth of 4 km is shown. From Figures 3.2 and 3.3 we can conclude that the
seismicity seems to be guided by the SE2 fault structure. A possible explana-
tion is that the SE2 fault system is characterized by an enhanced permeability
due to the presence of natural fractures. Such permeability is anisotropic. Pore-
pressure fluctuations due to the fluid injection then propagate mainly along the
direction of the largest principal component of the permeability tensor. This
component is directed along the fractures, i.e. along the fault. This leads up
to the phenomenon of fault-guided induced seismicity. Moreover, Figures 3.2
and 3.3 indicate that the permeability along the SE2 fault into the horizon-
tal (strike) direction is larger than the permeability along its up-dip direction.
This is possibly related to the geometry of fractures composing the SE2 fault
system.
In spite of only a small number of located events, the fluid-induced seismicity at
the KTB site shows several important features of events triggering by pore-pressure
perturbations. First, it was observed that seismicity can be triggered by pressure
perturbations as low as 0.01–0.1 MPa or possibly even less. This is seen from the
following. Most of the microseismic events occurred several hundred meters away
from the injection borehole (see Figures 3.2 and 3.3). Simultaneously the water
level in the main hole at the start of seismic activity had increased by a few tens of
Figure 3.2 A vertical slice of a 3D depth migrated image of the KTB site (see
Buske, 1999). Light tones correspond to high seismic reflection intensities and
dark tones to lower ones, respectively. The SE1 reflector is clearly visible as a
steeply dipping structure. Additionally, seismicity induced by the injection exper-
iments of years 1994 (close to the injection depth of 9.1 km), 2000 (injection
depth of approximately 5.6 km) and 2004–2005 (depth around 4 km) is shown.
Locations of the main (black line) and pilot (light-tone line) boreholes are also
plotted. (Modified after Shapiro et al., 2006a.) A black and white version of this
figure will appear in some formats. For the color version, please refer to the plate
section.
meters. Therefore pore-pressure perturbations less than few 0.1 MPa at hypocenters
are required to trigger seismic events at the KTB site. Second, seismicity trigger-
ing is much more likely in the case of positive pore-pressure perturbation (i.e. in
the case of an injection). For example, the probability of event triggering by the
fluid extraction at KTB was less than 1/3000 of the probability during fluid injec-
tion. No events occurred during the fluid extraction and more than 3000 occurred
during the injection. Moreover, the onset of seismicity roughly coincides with the
time of compensation of the extracted fluid volume by the following injection. This
confirms that pressure diffusion is the dominant mechanism of seismicity trigger-
ing by fluid injections. Finally, it was shown that heterogeneity and anisotropy
of hydraulic properties of rocks can influence spatial evolution of microseismic
Figure 3.3 A horizontal slice at 4 km depth over the depth-migrated 3D image of

the KTB site (Buske, 1999) plotted along with the slice-plane projections of seis-
mic hypocenters induced by the injection experiment of 2004–2005. The white
and gray squares are locations of the main and pilot boreholes, respectively. (After
Shapiro et al., 2006a.) A black and white version of this figure will appear in some
formats. For the color version, please refer to the plate section.
clouds. For example, the seismicity induced by the fluid injection directly into the
SE2 structure remained guided by this crustal fault.
3.2 Linear relaxation of pore pressure as a triggering mechanism

A heterogeneous distribution of elastic stresses is a consequence of a heterogeneity
of rock elastic properties (Langenbruch and Shapiro, 2014). At some locations in
the Earth’s crust, the tectonic stress is close to the critical stress necessary for a
brittle failure of rocks. Formally this means that the failure criterion stress FC S
given by equation (1.73) is still negative. However, it has small absolute values.
Increasing fluid pressure in a reservoir also leads to an increase of the pore pres-
sure at critical locations (i.e. at rock defects such as pre-existing cracks). Such
an increase in the pore pressure causes a decrease in the effective normal stress
(poroelastic-coupling stress contributions considered in Section 2.8 are neglected
in the first approximation; we will discuss their role again in Section 3.4). At some
locations the decrease in the effective normal stress produces a sufficiently high
positive change of FC S. This leads to sliding along pre-existing favorably ori-
ented cracks (sub-critical before the pore-pressure rise). This mechanism of event
triggering has been considered by, for example, Nur and Booker (1972), Fletcher
and Sykes (1977), Ohtake (1974), Talwani and Acree (1985), Ferreira et al. (1995)
and Zoback and Harjes (1997).
The linear relaxation of stress and pore pressure is described by the system
of Frenkel–Biot equations for small linear deformations of poroelastic systems
3.2 Linear relaxation of pore pressure as a triggering mechanism 125
3000 m
(a) 3600 m (b) (c) (d)
Figure 3.4 Hypocenters of microseismic events occurred during (a) 100 h, (b)
200 h, (c) 300 h and (d) 400 h after starting a water injection in a borehole of
the geothermal system at Soultz, France (after Shapiro et al., 1999; see also
Section 3.8.2 for a description of the case study).
(Biot, 1962). We discussed this system of equation in the previous chapter. This
equation system shows that, in a homogeneous isotropic fluid-saturated poroelastic
medium, there are three waves propagating a strain perturbation from a source to a
point of observation. These are two elastic body waves, P and S (the longitudinal
and shear seismic body waves, respectively), and a highly dissipative longitudinal
slow wave.
Clouds of fluid-induced microseismic events grow slowly. They need hours
or days to become several hundred meters long (see Figure 3.4). This process
cannot be explained by the elastic stress transfer produced mainly by quick seis-
mic body waves. Thus, the growth of microseismic clouds has to be related
to the slow wave. The pore-pressure perturbation in the slow wave in the
limit of frequencies extremely low in comparison with the global-flow critical
frequency (which is usually of the order of 0.1–100 MHz for realistic geo-
logic materials) is described by a linear partial-differential equation of diffusion
(2.114). It is exactly the same diffusion equation that can be obtained by uncou-
pling the pore pressure from the complete Frenkel–Biot equation system in the
low-frequency range. In addition, the uncoupling of the pore-pressure diffusion
equation requires an assumption of an irrotational displacement field in the solid
skeleton (Detournay and Cheng, 1993). In weakly heterogeneous and weakly
elastically anisotropic rocks far from high-contrast poroelastic boundaries, this
assumption is approximately valid. Then the diffusion equation describes well
enough the linear relaxation of pore-pressure perturbations in a poroelastic fluid
saturated medium.
Shapiro et al. (1999, 2002, 2003) proposed using the diffusion equation (2.120)
to describe the spatio-temporal evolution of clouds of hydraulically induced
seismic events in terms of pore-pressure relaxation in anisotropic and heteroge-

neous porous media:

∂ Pp ∂ ∂
= Di j Pp . (3.1)
∂t ∂ xi ∂x j
Here, Di j are the components of the tensor of hydraulic diffusivity, x j ( j = 1, 2, 3)

are the components of the radius vector from the injection point to an observation
point in the medium, Pp is the pore-pressure perturbation with respect to the pre-
injection in situ pore pressure, and t is time.
Under the assumptions mentioned above, by only allowing additionally for
heterogeneity and anisotropy of the hydraulic permeability, (3.1) can still be uncou-
pled from the complete system of the Frenkel–Biot equations in the low-frequency
range (see Section 2.4.4). The coefficient D here is the same coefficient of hydraulic
diffusivity derived by Rice and Cleary (1976) and Van der Kamp and Gale (1983)
in the notation of Biot (1956). It has the following relation to the permeability
tensor (see equation (2.119)):
k
D= , (3.2)
ηS
where k is the tensor of hydraulic permeability, η is the pore-fluid dynamic

viscosity and S is the storage coefficient (2.100):
K dr + 43 μdr + α 2 M
S= , (3.3)
M(K dr + 43 μdr )
where M = (φ/K f + (α − φ)/K gr )−1 and α = 1 − K dr /K gr . Here K f,dr,gr are

bulk moduli of the fluid, dry frame and grain material, respectively; μdr is the shear
modulus of the frame and φ is the porosity.
Note, again, that here we neglected the elastic anisotropy in comparison with
the anisotropy of the permeability. It is clear that, in reality, anisotropy and het-
erogeneity of the permeability are related to anisotropy and heterogeneity of
elastic properties of rocks, at least on the micro-scale. However, usually the elastic
anisotropy has the order of 5%–30%. On the other hand, the permeability has an
anisotropy of the order of several hundred percent or more. Thus, in equation (3.2),
elastic anisotropy can be neglected for simplicity.
For the case of highly porous rocks, an approximation
φ
S≈ (3.4)
Kf
can be used to roughly estimate this quantity. For the case of low-porosity
crystalline rocks an order-of-magnitude approximation of S is:
φ α
S≈ + . (3.5)
Kf K gr
In the following discussion we use the diffusion equation in order to construct
an approximate model of the process of microseismicity triggering by borehole
fluid injections. Of course, it would be possible to introduce the diffusion equa-
tion without any relation to the slow wave. However, understanding of this relation
provides us with several advantages. First, we are able to derive a relationship
between the hydraulic diffusivity and other poroelastic parameters in terms of
physical quantities used in exploration seismology. Second, we understand that the
diffusion equation describes the phenomenon in the low-frequency approximation
only. For example, this explains an unphysical result of this equation: an infinitely
quick propagation of high-frequency components of pore-pressure perturbations.
In reality, the slow wave always has a finite (or zero) propagation velocity. Finally,
we understand that the diffusion equation in its classical simple form is valid under
conditions of pore-pressure uncoupling from the elastic stress only. These condi-
tions are, in turn, related to the low-frequency range under consideration. In the
previous chapter we discussed the poroelastic coupling effects. In Section 3.4 we
will return once more to this subject.
3.3 Triggering front of seismicity

The spatio-temporal features of the hydraulically induced seismicity can be iden-
tified in a very natural way from the concept of triggering fronts (Shapiro
et al., 1997, 2002). For the sake of simplicity we approximate a real configuration
of a fluid injection in a borehole by a point source of pore-pressure perturbation
in an infinite, hydraulically homogeneous and isotropic poroelastic fluid-saturated
medium. The time evolution of the pore pressure at the injection point is taken to
be a step function switched on at time 0. It is natural to assume that the probability
of the triggering of microseismic events is an increasing function of the magnitude
of the pore-pressure perturbation. Thus, at a given time t it is probable that events
will occur at distances that are smaller than, or equal to, the size of the relaxation
zone (i.e. a spatial domain of significant changes) of the pore pressure. The events
are characterized by a significantly lower occurrence probability for larger dis-
tances. The spatial surface that separates these two spatial domains will be called
the “triggering front.”
Because the size of the relaxation zone is a rather heuristic parameter, we intro-
duce a more formal quantity, which is directly proportional to this size. At time t,
1
0.8
0.6
0.4
0.2
0
0 1 2 3 4
ω t
2p
Figure 3.5 The power spectrum of a boxcar-function signal.
the boxcar-function signal of the duration t produces pore-pressure perturbations

at any point of the medium. The frequency 2π/t is an upper bound of the dominant
frequencies of this boxcar signal (see Figure 3.5). We define the size of the relax-
ation zone (i.e. the relaxation radius) to be equal to a distance traveled in a time
t by the phase front of a harmonic pore-pressure diffusion wave of the frequency
2π/t. Note that lower-frequency pressure-diffusion waves propagate with lower
velocities.
In a homogeneous and isotropic medium this definition results in the following
equation of the triggering front (we use equation (2.103) for the phase velocity of
the slow wave):
√
Rt = 4π Dt, (3.6)
where t is the time from the injection start, D is a scalar hydraulic diffusivity and Rt
is the radius of the triggering front (which is a sphere in a homogeneous isotropic
medium). Note that the numeric coefficient 4π in (3.6) is still heuristic. Defined in
this way, the relaxation radius is equal to the wavelength of the slow wave with the
period equal to t (see equation (2.104)). Geological media are usually hydraulically
heterogeneous. Equation (3.6) is then an equation for the triggering front in an
effective (replacing) isotropic homogeneous poroelastic medium with the scalar
hydraulic diffusivity D.
Because a seismic event is much more probable in the relaxation zone than at
larger distances, equation (3.6) corresponds approximately to the upper bound of
the cloud of events in a plot of their spatio-temporal distribution (i.e. the plot of
r versus t called in the following discussion the r –t plot). We will first demon-
strate this by using the following simple numerical experiment. A point source
of a fluid injection is embedded into an infinite homogeneous porous continuum.
As a result of a fluid injection and the consequent process of pressure relaxation,

pore pressure will change throughout the pore space. A step-function pressure per-
turbation is used as an input injection signal. We solve the diffusion equation for
the pore-pressure perturbation Pp at any point of the model at any time. For a
hydraulically heterogeneous model one can use a finite element algorithm (see
Rothert and Shapiro, 2003, for more-complex model examples). For a homoge-
neous isotropic medium solutions (2.196) and (2.164) can be used directly. After
time t0 the injection is stopped. The medium is divided into small cells. We assume
that a critical value C of the pore-pressure perturbation is randomly distributed in
the medium.
If at a given point r of the medium at a given time t the pore pressure Pp (t, r)
exceeds C(r) then this point will be considered as a hypocenter of an earthquake
that occurred at the time t. For simplicity, we assume that no earthquake will be
possible at this point again. We address the quantity C as criticality.
Figures 3.6–3.8 show the distribution of the criticality C and resulting synthetic
clouds of events generated by such numerical experiments on rock stimulations
by fluid injections in two hydraulically identical homogeneous models with the
hydraulic diffusivity D = 1 m2 /s. These models differ only in the spatial distribu-
tion of the critical pore pressure C. Figure 3.8 shows distances r from the injection
points (0, 0) located at the centers of the models to points with simulated seismic
events versus event-occurrence times t. In our terminology, this figure shows the
r –t plots. It is evident that a vast majority of points corresponding to the seismic
events is located below approximately parabolic envelopes. Curves (3.6) with the
50 50
40 40
30 30
20 20
10 10
0 0
–10 –10
–20 –20
–30 –30
–40 –40
–50 –50
–50 –40 –30 –20 –10 0 10 20 30 40 50 –50 –40 –30 –20 –10 0 10 20 30 40 50
Distance [m] Distance [m]
Figure 3.6 Left: realizations of the critical pore pressure C randomly distributed
in space with an exponential auto-correlation function. Right: the same but with a
Gaussian auto-correlation function. (Modified from Rothert and Shapiro, 2003.)
A black and white version of this figure will appear in some formats. For the color
version, please refer to the plate section.
Events and occurence time Events and occurence time
50 50 100
40 40 90
30 30 80
20 20 70
Distance [m]
10 10 60
Distance [m]
0 0 50
–10 –10 40
–20 –20 30
–30 –30 20
–40 –40 10
–50 –50
–50 –40 –30 –20 –10 0 10 20 30 40 50 –50 –40 –30 –20 –10 0 10 20 30 40 50
Figure 3.7 Synthetic microseismicity simulated for the two spatial distributions
of the critical pore pressure shown in Figure 3.6, respectively. (Modified from
Rothert and Shapiro, 2003.) A black and white version of this figure will appear
in some formats. For the color version, please refer to the plate section.
50 50
45 45
40 40
35 35
Distance [m]
Distance [m]
30 30
25 25
20 20
15 15
10 10
5 5
0 0
0 10 20 30 40 50 60 70 80 90 100 0 10 20 30 40 50 60 70 80 90 100
Time [s] Time [s]
Figure 3.8 Plots of distance versus times (r –t plots) for the synthetic micro-
seismic clouds shown in Figure 3.7, respectively. The dots are events. The lines
correspond to D = 1 m2 /s. (Modified from Rothert and Shapiro, 2003.)
hydraulic diffusivity D = 1 m2 /s are also shown in Figure 3.8. It is obvious that

the spatio-temporal distribution of events agrees well with the behavior described
by equation (3.6). Let us assume that we do not know the hydraulic diffusivity
of the medium. In this case equation (3.6) can be used to estimate the diffusivity.
Thus, a parabolic-like envelope of the cloud of events in an r –t plot is an important
signature of the hydraulically triggered seismicity.
Such a spatio-temporal distribution of the microseismicity is often observed for
microseismic data in reality. Two examples of fluid-injection-induced microseis-
micity in crystalline rocks are shown in Figure 3.9. We see that r –t plots provide
us with estimates of hydraulic diffusivity. We will address them later in this chapter.
(a) (b)
1000
Events Fenton Hill 1000
Distance from injection source [m] Events Soultz ’93
Distance from injection source [m]

900 900
D = 0.17 m2/s D = 0.05 m2/s
800 800
700 700
600 600
500 500
400 400
300 300
200 200
100 100
0 0
0 10 20 30 40 50 0 50 100 150 200 250 300 350 400
Time from beginning of injection [h] Time from beginning of injection [h]
Figure 3.9 The r –t plots for microseismic clouds induced by two water injection
experiments of two different Enhanced Geothermal System projects. One is at
Fenton Hill, USA (a) and another is at Soultz, France (b). More details on the case
studies are given in Sections 3.8.2 and 3.8.3. (Modified from Shapiro et al., 2002.)
3.4 Seismicity fronts and poroelastic coupling

Let us consider further the concept of the triggering front from the point of view
of a perturbation of the poroelastic failure criterion stress. Such perturbations are
expressed by equations (2.236) and (2.237) of the previous chapter. They show that
poroelastic contributions to the failure criterion stress depend on the dimension-
less functions erfc(r0 ), sdi f (r0 ) and sm (r0 ) of the normalized distance r0 defined
in (2.160). They depend also on the coupling parameter n s and on the friction
angle φ f . In such a criterion, different stress functions contribute with different
magnitudes.
To compare these contributions let us consider Figure 3.10. This figure pro-
vides plots of changes of the failure criterion stress (2.236)–(2.237) corrected for
geometrical spreading and normalized to the source term:
r FC Sh
δ FC Sh (r0 ) = , (3.7)
Cf
r FC Sv
δ FC Sv (r0 ) = . (3.8)
Cf
They are denoted in Figure 3.10 by the dashed lines and dotted lines, respec-
tively. For these plots the following parameters have been used: the friction angle
φ f = 30◦ and the poroelastic stress coefficient n s = 0.3 for the left-hand part of
Figure 3.10 and n s = 0.05 for the right-hand part of Figure 3.10. The solid curves
in Figure 3.10 show functions
0.5 0.5
0.4 0.4
0.3
0.3
0.2
0.2
0.1
0.1
0
0.01 0.05 0.1 0.5 1 5 10 0
r0
0.01 0.05 0.1 0.5 1 5 10
r0
Figure 3.10 Normalized perturbations of failure criterion stress δ FC S(r0 ) caused

by a constant-rate fluid injection and taking into account poroelastic coupling
effects. The left-hand part shows a very strong poroelastic coupling with n s = 0.3.
The right-hand part shows a weak poroelastic coupling with n s = 0.05. By fluid
stimulations of extremely low-permeable tight rocks, even weaker coupling is
expected (see Figure 2.11). Dotted curves correspond to locations along the com-
pressive stress with the maximum absolute value. These are functions δ FC Sv (r0 ).
Dashed curves correspond to locations along the compressive stress with the min-
imum absolute value. These are functions δ FC Sh (r0 ). Solid curves correspond
to the effect of the pore-pressure perturbation only, δ FC S p (r0 ). The values of
δ FC S(r0 ) relevant for fitting domains of seismicity on r –t plots in practical situa-
tions are close to or above 0.1 (the horizontal solid lines given by equation (3.11)).
For example, this corresponds to ratios r/a0 > 30 (a0 is the radius of the injection
source) and Cs / p0 > 0.003 (see Figure 3.9). The vertical solid lines correspond
to the triggering front defined by (3.6)
r FC S p (r0 )
δ FC S p (r0 ) = = erfc(r0 ) sin φ f . (3.9)
Cf
These curves describe a pure pore-pressure effect onto the failure criterion stress.
We consider the case of a constant-rate (or a constant-pressure) fluid injection.
This means that C f is positive (see equations (2.220) and (2.221)). The dotted
curves in Figure 3.10 correspond to the maximum change of the FC S. One could
define a triggering front as a solution of the following equation:
Cs = FC Smax (r, t), (3.10)
where Cs is a finite stress perturbation (close to the minimum criticality C). In the
case of a fluid-injection, equation (3.10) simplifies to the following one:
rCs
δ FC Sv (r0 ) = . (3.11)
Cf
A solution of this equation will provide r as a function of t. This function will

be a proxy of the triggering front defined by a constant effective-stress perturba-
tion assumed to be necessary for inducing seismic events. We will denote such a
front estimate as Rc (t). Usually very early microseismic events, often located very
close to the injection source, are not considered for further interpretation. Such a
restriction to the distance from the injection source to events is frequently nec-
essary due to a high probability of event identification and location errors in the
close vicinity of the injection source. Let us consider a range of distances above
r = 30 m. This is a reasonable restriction for nearly all case studies from our book.
At these distance ranges, FC Smax (r, t) corresponds in Figure 3.10 to the dotted
lines in the domains close to and above the solid horizontal lines. These lines are
given by the right-hand side of equation (3.11) for parameters Cs = 0.03 MPa,
p0 = 10 MPa and a0 = 1 m. The function Rc (t) is given by the values of the argu-
ment r0 of the dotted curves in the domains above the horizontal lines. The vertical
lines correspond to the triggering front (3.6). This vertical line constrains quite
well the zone of significantly perturbed pore pressure (solid curves). It still pro-
vides quite reasonable constraints even for the failure criterion stress perturbations
taking the maximum and minimum poroelastic coupling contribution into account
(the dotted and dashed curves). The triggering-front equation (3.6) describes the
size of a spatial domain of significant perturbations of the pore pressure around the
injection source. This size was assumed to be equal to the wavelength of the pore-
pressure diffusional wave (corresponding to Biot’s slow P-wave) of period t. Such
a definition of the triggering front does not use any information on the hypothetic
quantity Cs .
√
Note that equation (3.6) can be written as r0 = π (see again equation (2.160)).
For any monotonic in time (non-decreasing) finite-stress perturbation proportional
to a function C f f (r0 )/r this equation will give a time instant at which this pertur-
√
bation reaches the f ( π )/ f (0)-part of its maximally possible value C f f (0)/r at
any given distance r from the source. Therefore, if the quantity Cs is not a constant
but rather it is heterogeneously distributed in rocks (for example, it takes values
from a broad range between nearly zero and several MPa) then a time dependence
of the triggering front will be given by an equation of the type of (3.6). This is
similar to a front of a spherical seismic wave (also having a 1/r -type of geometric
spreading), which is given not by a constant amplitude but rather by a constant
√
argument of the phase function. In other words the t-dependence will be a fea-
ture of the triggering front independently of the nature of the stress perturbation:
pure pore pressure or poroelastic stress coupled with the pore pressure. Such a
temporal dependence is a universal feature of a linear diffusion. It is applicable
under the assumption that no significant changes of the hydraulic diffusivity D
occur in the medium during the fluid injection. In the next chapter we will see
that stress perturbations caused by a non-linear diffusion can deviate from the
√
t-dependence.
The quantity Rc (t) defined by (3.11) and a constant Cs does not have a universal
√
t-dependence. Moreover, even in the case of a complete absence of poroelastic
effects (i.e. pure pore-pressure-caused FC S perturbations) Rc (t) does not have a
√
universal t-dependence. The smaller the value of Cs takes, the larger will be
the difference between Rc and Rt . For small Cs or large r0 one can use a series
expansion of (3.11) and find an approximate solution:

C f Dn s (3 − sin(φ f ))t 1/3
Rc (t) = . (3.12)
Cs
Therefore, Rc depends not only on the diffusivity of the medium D but also on the
poroelastic coupling parameter n s . Moreover, the parameter combination Dn s /C,
if kept constant, will provide the same function Rc (t) for an infinite number of
different poroelastic media.
In respect to the triggering front we conclude the following. The poroelastic
coupling can influence the shape of the triggering front of induced seismicity (see
also Section 2.8). This influence depends strongly on the coupling parameter n s .
If n s becomes vanishingly small, all curves of Figure 3.10 will coincide. If in
a data-fitting domain these curves are close, then poroelastic coupling contribu-
tions are not very significant and the pore-pressure contribution can be used as a
good approximation. In such a case, the triggering front Rt (vertical line in Figure
3.10) and equation (3.6) provide a reasonable estimate of D. Such an estimate is
independent of any choice of the hypothetic quantity Cs .
A triggering front defined by an isobar Cs of the pore-pressure perturbation after
injection termination can also be fitted to microseismic data (see Edelman and
Shapiro, 2004, and Langenbruch and Shapiro, 2010).
3.5 Seismicity and hydraulic anisotropy

Hydraulic properties of rocks are often characterized by a strong anisotropy. Let
us now assume that components of the tensor of hydraulic diffusivity Di j are
homogeneously distributed in the medium. Under such an assumption we replace
a heterogeneous, seismically active rock volume by an effectively homogeneous,
anisotropic, poroelastic, fluid-saturated medium with an upscaled permeability
tensor. By using equation (3.1) for homogeneous anisotropic media, Shapiro
et al. (1999) showed how to generalize equation (3.6).
Indeed, assuming a homogeneous anisotropic medium, the equation of diffu-
sion can be written in the principal coordinate system of the tensor of hydraulic
diffusivity:
3.5 Seismicity and hydraulic anisotropy 135
∂ Pp ∂2 ∂2 ∂2
= D11 2 Pp + D22 2 Pp + D33 2 Pp , (3.13)
∂t ∂ x1 ∂ x2 ∂ x3
where D11 , D22 and D33 are the principal components of the diffusivity tensor and
xi are the spatial coordinates of points in the corresponding Cartesian system.
Let us consider the following new system of coordinates, X i (note summation
over repeating indices):
Dii Dii Dii

X 1 = x1 , X 2 = x2 , X 3 = x3 . (3.14)
3D11 3D22 3D33
In this system, equation (3.13) becomes
2
∂ Pp Dii ∂ ∂2 ∂2
= + + Pp . (3.15)
∂t 3 ∂ X 12 ∂ X 22 ∂ X 32
This equation is equivalent to equation (2.114); that is, it describes isotropic
relaxation of the pore pressure with the isotropic hydraulic diffusivity Dii /3. Thus,
in the new coordinate system X i , equation (3.6) keeps its form. The radius of the
triggering front is given by equation:
Dii
Xi Xi = 4π t. (3.16)
3
The left-hand side of this equation is a radius in the new coordinate system. Taking
the square of the radius and returning to the principal coordinate system xi , we
obtain
x12 x2 x2
+ 2 + 3 = 4πt. (3.17)
D11 D22 D33
On the other hand it is easy to see that in the principal coordinate system of the
diffusivity tensor (the matrix Di j is diagonal) the following equivalence is valid:
x12 x2 x2
+ 2 + 3 = xi [D −1 ]ik xk , (3.18)
D11 D22 D33
where [D −1 ]ik are components of the tensor inverse to the hydraulic-diffusivity
tensor in the principal coordinate system.
Thus, the right-hand sides of equations (3.17) and (3.18) must be equivalent.
They can be written in a coordinate-independent form:
rT D−1 r = 4πt, (3.19)
where the superscript T denotes that the matrix (vector) is transposed, r is a radius
vector of a point on the triggering front and D−1 is the tensor inverse of D.
Equation (3.19) replaces equation (3.6) in the more general case of a hydrauli-
cally homogeneous anisotropic medium. This equation holds its form in an
arbitrary rotated (i.e. not necessary principal) coordinate system. Written in a sim-
pler form, characteristic for the principal coordinate system (see equation (3.17)),
it shows that the triggering front is an ellipsoidal surface. If we scale the principal
coordinate system as
√
xs j = x j / 4πt, (3.20)
then the equation of the triggering front will become an equation of an ellipsoid
with the half-axes equal to the square roots of the principal diffusivities. Therefore,
if we scale the coordinates of all events by the square root of their occurrence time,
by analogy with equation (3.6) this ellipsoid will be an envelope of the cloud of
events, but now in a 3D space, normalized accordingly with (3.20). Such an ellip-
soidal envelope (see Figure 3.11) can be used to estimate spatial orientations and
values of principal components of the hydraulic diffusivity. Shapiro et al. (2003)
demonstrated applications of this approach to different case histories. For example,
in the case of the Fenton Hill experiment (Figure 3.11 and Section 3.8.3) they esti-
mated anisotropic tensor of hydraulic diffusivity with an approximate relation of
the principal components of 1:4:9. In the case of the Soultz experiment (see Sec-
tion 3.8.2 for more details) a tensor of hydraulic diffusivity with an approximate
relation of the principal components of 1:2.5:7.5 was obtained. Corresponding
estimates of maximal principal components of the diffusivity are 0.17 m2 /s and
0.05 m2 /s, respectively (see Figure 3.9).
Shapiro et al. (1999) described a slightly different approach for estimating the
diffusivity tensor. They proposed dividing the entire space into directional sectors
centered at the injection source. Then, they evaluated r –t plots in these sectors
independently. Parabolic envelopes of seismicity at the r –t plots represent corre-
sponding sections of an ellipsoidal triggering front growing with time in space. A
set of diffusivity estimates obtained in the sectors can be further used to reconstruct
the complete 3 × 3 diffusivity matrix.
The diffusivity estimates shown above also provide us with a possibility to assess
the hydraulic permeability of rock. For this, equation (3.2) can be used:
k = DSη. (3.21)
For the Fenton Hill experiment, we accept the following estimates used in the liter-
ature for crystalline rocks at the depth of 3500 m: φ = 0.003, η = 1.9 · 10−4 Pa·s.
(dynamic viscosity of salt water at 150 ◦ C), K d = 49 GPa, K g = 75 GPa and K f =
2.2 GPa. From these values we obtain S −1 ≈ 1.68 · 1011 Pa, and the permeability
tensor in the principal coordinate system is
k = diag (0.2 ; 0.8 ; 1.8) · 10−16 m2 . (3.22)

Figure 3.11 Top: a microseismic cloud (Fenton Hill, see Section 3.8.3 for more
details about the case study) shown in three different projections in conventional
coordinates. Middle: the same cloud projections in coordinates normalized corre-
spondingly to equation (3.20). Bottom: the same as before, but now with a fitted
ellipsoid. (Modified from Shapiro et al., 2003.)
Approximate estimates of the permeability at Fenton Hill by other authors give

10−17 –10−15 m2 (see Pearson, 1981).
The microseismic experiment in Soultz mentioned above provided the following
estimates (Shapiro et al., 1999):
k = diag (0.7 ; 1.9 ; 5.2) · 10−17 m2 . (3.23)
Determinations of the apparent permeability from independent hydraulic data (see
Jung et al., 1996) provided the estimate of 2.5 · 10−17 m2 .
3.6 Seismicity in hydraulically heterogeneous media

The equations given above are able to provide scalar or tensorial estimates of the
hydraulic diffusivity in an effective homogeneous medium only. Shapiro (2000)
2000
1000
Distance [m]
0
−1000
−2000
−2000 −1000 0 1000 2000
Distance [m]
Figure 3.12 A numerical example of pore-pressure distribution (both the gray

scale and the isolines) from a point source of an injection located in a layer
with hydraulic diffusivity smaller by a factor of five than that of the surrounding
medium. Note that the isolines of the pore pressure are similar to fronts of prop-
agating waves. Figure courtesy of Elmar Rothert. (From Shapiro, 2008, EAGE
Publications bv.)
and Shapiro et al., (2002) proposed an approach to map 3D heterogeneously dis-

tributed hydraulic diffusivity. They heuristically used an analogy between pressure
diffusion and wave propagation.
Figure 3.12 shows an example of the pore-pressure distribution created at a
given time by a point source of a constant strength acting in a 2D numerical model.
The source of pressure perturbation is located in a layer of low hydraulic diffu-
sivity. This layer is embedded into a medium with a five-times-larger hydraulic
diffusivity. Figure 3.12 clearly illustrates similarities of the pressure isolines with
fronts of propagating waves.
By using the analogy between diffusion and wave propagation it is possi-
ble to derive a differential equation that approximately describes the kinematics
of the triggering front in the case of quasi-harmonic pore-pressure pertur-
bation. This approximation is similar to the geometrical-optics approach for
seismic waves. Propagation of the triggering front is considered in a lim-
ited frequency range, which can be called an intermediate asymptotic one in
the following sense. The dominant frequency of the injection-induced pres-
sure perturbations must be much smaller than the critical Biot frequency. On
the other hand it is assumed that the dominant wavelength of the slow wave
(i.e. the pore-pressure diffusion wave) is smaller than the characteristic size
of the heterogeneity of the hydraulic diffusivity. Under these conditions one
can show that arrival times of the triggering front are described by an eikonal
equation.
3.6.1 Eikonal-equation approach

Let us consider the relaxation of a time-harmonic pore-pressure perturbation. First
we recall the form of the solution of equation (3.1) in the case of a homogeneous
isotropic poroelastic medium (see, for example, chapter V of Landau and Lifshitz,
1991). If a time-harmonic perturbation p0 exp(iωt) of the pore pressure is given
on a spherical surface of the radius a0 with its center at the injection point, then the
solution is

iωt a0 ω
Pp (r, t) = p0 e exp −(i + 1)(r − a0 ) , (3.24)
r 2D
where ω is the angular frequency and r is the distance from the injection point
(i.e. the center of the sphere) to the point where the solution is sought. Comparing
this result with equations (2.102) we note that equation (3.24) describes a spheri-
√
cal slow compressional √ wave with attenuation coefficient equal to ω/(2D) and
slowness equal to 1/ 2Dω.
By analogy to solution (3.24) we look for solutions of equation (3.1) of the
following form:
√
Pp (r, t) = p0 (r)eiωt exp −τ p (r) ω . (3.25)
We assume that p0 (r), τ p (r), and Di j (r) are frequency-independent functions

slowly changing with r (later in this section we will quantify this condition). Sub-
stituting equation (3.25) into equation (3.1), accepting ω as a large parameter, and
keeping only terms with the largest exponents (these are terms on the order of
√
O(ω); other terms, on the orders of O(ω0 ) and O( ω), are neglected), we obtain
∂τ p (r) ∂τ p (r)
i = Di j (r) . (3.26)
∂ xi ∂x j
On the other hand, by analogy to equations (2.102) and (3.24) the frequency-
independent quantity τ p is related to the frequency-dependent phase travel time T p
(see the comment after equation (3.30)) as follows:
√
τ p = (i + 1)T p ω. (3.27)
Substituting this into equation (3.26), we obtain

∂ T p (r) ∂ T p (r)
1 = 2ωDi j (r) . (3.28)
∂ xi ∂x j
In the case of an isotropic poroelastic medium, Di j (r) = δi j D(r) and equa-
tion (3.28) reduces to
1
|∇T p (r)|2 = . (3.29)
2ωD(r)
This is a so-called eikonal equation. The right-hand side of this equation is the
squared phase slowness of a wave (the slow wave in this particular case). One
can show (Cerveny, 2005) that eikonal equations are equivalent to Fermat’s prin-
ciple, which ensures minimum-time (stationary-time) signal propagation between
two points of a medium. Because of equation (3.27), the minimum travel time
corresponds to the minimum signal attenuation of the slow wave. In this sense,
equations (3.28) and (3.29) describe the minimum-time, maximum-energy front
configuration.
Let us further consider a more realistic situation, where the pressure perturbation
at a point source can be roughly approximated by a step function. Equations (3.28)
and (3.29) describe the phase travel time T p of a harmonic pressure perturbation.
We can attempt to use these eikonal equations further to derive an equation for the
triggering time t of a step-function pressure perturbation. For this we can express
T p for an arbitrary frequency ω in terms of the triggering time t.
From our earlier discussion we know that the triggering time t roughly
corresponds to the frequency
2π
ω0 = . (3.30)
t
Thus, for the particular frequency ω = ω0 we can roughly approximate the time
T p of the corresponding harmonic pressure perturbation by t. On the other hand,
√
we know that generally T p (ω) ∝ 1/ ω. Therefore, if the phase travel time T p0 is
known for a particular frequency ω0 , then for another frequency ω (from a narrow
frequency range around ω0 ) the corresponding time T p will be given by T p =
√
T p0 ω0 /ω. Thus, for an arbitrary frequency ω, we obtain
2πt
Tp = . (3.31)
ω
Substituting this equation into equations (3.28) and (3.29) we obtain the fol-
lowing results. In the general case of an anisotropic heterogeneous poroelastic
medium,
∂t (r) ∂t (r)
t = π Di j (r) . (3.32)
∂ xi ∂ x j
In the case of an isotropic poroelastic medium, this equation reduces to
t
D(r) = . (3.33)
π |∇t|2
In the case of a homogeneous medium this equation is in agreement with the defi-
nition of the triggering front (3.6). Expressing the triggering time as a function of
the distance by using (3.6) and substituting the triggering time into (3.33) leads to
the identity D = D.
It follows from the previous discussions that equations (3.32) and (3.33) are
limited to low-frequency diffusion-type Biot slow waves. However, they can be
helpful for the characterization of hydraulic properties of rocks. In the case of an
isotropic poroelastic medium, equation (3.33) can be used directly to reconstruct a
3D heterogeneous field of hydraulic diffusivity. In turn, by using equation (3.32)
for an anisotropic medium, it is impossible to reconstruct a 3D distribution of the
diffusivity tensor (because this is a single equation with six unknowns). Here addi-
tional information is required. For example, if one assumes that the orientation
and the principal-component proportions are constant, then the tensor of hydraulic
diffusivity can be expressed as Di j (r) = D(r)di j . Here di j is a non-dimensional
constant tensor of the same orientation and principal-component proportion as the
diffusivity tensor and D is the heterogeneously distributed magnitude of this ten-
sor. If the tensor di j is known (for example, it can be estimated from the symmetry
of the microseismic cloud, as discussed in the previous section), then D can be
computed directly as

∂t (r) ∂t (r) −1
D(r) = t π di j . (3.34)
∂ xi ∂ x j
If di j is normalized so that dii = 3, then in the case of an isotropic medium it is
equal to the unit matrix.
Figure 3.13 shows an example of a reconstructed spatial distribution of the
hydraulic diffusivity obtained for the Soultz data set according to equation (3.33).
Assuming that di j has the same orientation and principal-component proportion as
permeability tensor (3.23), equation (3.34) can be applied to obtain the diffusivity
tensor magnitude. Without showing this in detail, it is interesting to note that there
is no significant difference between results of the isotropic and anisotropic variants
of the method. Both show larger diffusivity in the upper part of the stimulated vol-
ume than in its lower part. In addition, a highly permeable channel leading to the
upper right-hand part of the structure is visible in the reconstructed hydraulic dif-
fusivity. This agrees with the observation of a number of early events in the upper
right-hand corner of the rock volume.
3.6.2 Validity domain of the eikonal-equation approach

The main limitations of the validity range of equation (3.26) can be formulated
roughly from the following consideration of the right-hand part of equation (3.1)
in a 1D medium:

∂ ∂ ∂ D ∂ Pp ∂ 2 Pp
D Pp = +D . (3.35)
∂x ∂x ∂x ∂x ∂x2
x
250 m y
0.001 500 m
z
0.01
0.1
1
1000 m
D (m2/s)
Figure 3.13 Reconstruction of the diffusivity distribution in a hydraulically het-

erogeneous geothermic reservoir of Soultz (corresponding to the stimulation of
1993) using the eikonal-equation approach. The dark-tone grid cells in the upper
part of the structure have hydraulic diffusivity in the range 0.1–1 m2 /s. Below
1000 m the structure has the diffusivity mainly in the range 0.001–0.05 m2 /s.
(Modified from Shapiro et al., 2002.) A black and white version of this figure will
appear in some formats. For the color version, please refer to the plate section.
Our approach is expected to be valid if the local heterogeneity of the medium can
be neglected. Thus, the following inequality must be satisfied:

∂ D ∂ Pp ∂ 2 Pp

∂ x ∂ x / D ∂ x2 1 (3.36)
This can be roughly reduced to

∂ D

∂ x /|Dκd | 1, (3.37)
where κd is the wavenumber of the slow wave. Taking into account that |κd |2 ≈
ω/D (see equation (2.98)), we obtain the following simplified condition:

∂ D 2 −1

∂ x D ω. (3.38)
This inequality relates a spatial gradient of the hydraulic diffusivity and the fre-
quency of the pressure perturbation. It is rather typical for the geometrical-optics
approximation. It shows that if the frequency is high enough and the medium
heterogeneity is smooth, the eikonal equation can be applied.
In the case of a step-function pressure perturbation, the frequency correspond-
ing to the triggering front is roughly given by (3.30). Using equation (3.6) of the
triggering front in homogeneous poroelastic media, we can approximate the occur-
rence time of early events as t ≈ x 2 /(4D). Note that x denotes the distance from
the injection source. Thus, inequality (3.38) can be approximately reduced to

x ∂D

D ∂ x < 3. (3.39)
This condition is rather restrictive. In addition, it shows that the smaller the
distance x, the higher the resolution of the method. In spite of the restrictive
character of the inequalities above, the geometrical-optics approximation is still
applicable to propagating microseismicity triggering fronts under rather common
conditions. This is based on the causal nature of the triggering-front definition.
When considering the triggering front, we are interested in the quickest possi-
ble configuration of the phase traveltime surface for a given frequency. Thus, we
are interested in kinematic aspects of the front propagation only. Corresponding
description is given by the eikonal equation. However, conditions (3.38) and (3.39)
necessarily take into account not only kinematic aspects of the front propagation
but also dynamic aspects, i.e., amplitudes of the pressure perturbation. In other
words, the eikonal equation is usually valid in a much broader domain of frequen-
cies than those given by inequality (3.38). Therefore, the method will give useful
results, at least semiquantitatively.
Shapiro et al. (2002) and Rothert and Shapiro (2003) performed some numer-
ical tests of applicability of the eikonal equations derived above to a kinematic
description of the diffusion process and the evolution of the triggering front in a het-
erogeneous media. They shown that the approach provides reasonable approximate
results for models with heterogeneous distributions of the hydraulic diffusivity in
rocks.
3.6.3 Effective-medium approach

Let us assume that a medium is heterogeneous on a small scale below the level
of our resolution. For example, we estimate the diffusivity in a heterogeneous
medium using just a microseismic triggering front at its final observed position
(sometimes, at the moment of the injection termination). In this case we attempt
to replace the real medium by an effective homogeneous one. Thus, we obtain an
effective (upscaled) diffusivity estimate. Below we derive some rules of computing
such an effective diffusivity in heterogeneous spherically symmetric d-dimensional
structures and discuss their applicability for interpretation of microseismic data.
We consider the system of continuity and Darcy equations (2.64) and (2.66),
respectively. The medium is assumed to be isotropic and heterogeneous. A point
fluid source is located at the origin of the coordinate system. Fluid filtration from
such a source corresponds to radiation of a slow wavefield (see Section 2.6.1).
Under such conditions, according to equation (2.113), the quantity χ , the vol-
umetric deformation of the pore space due to additional fluid-mass filtration, is
proportional to the pore-pressure perturbation Pp . Then, in a spherically symmetric
d-dimensional medium, diffusion equation (3.1) is obtained from the following

equation system:
∂ Pp 1 ∂r d−1 qr
S(r ) = − d−1 , (3.40)
∂t r ∂r
∂ Pp
qr = −D(r )S(r ) , (3.41)
∂r
where r is a radial distance, qr is the radial filtration velocity, and S(r ) and D(r ) are
the radial-distance-dependent storage coefficient and hydraulic diffusivity, respec-
tively. It is clear that, by substituting the filtration velocity from the second equation
into the first one, we obtain the diffusion equation
∂r d−1 Pp 1 ∂ ∂
= S Dr d−1 Pp . (3.42)
∂t S ∂r ∂r
Let us first consider steady-state filtration. From (3.40) we obtain
∂r d−1 qr
= 0. (3.43)
∂r
This corresponds to a system having an infinitely long time to equilibrate. From
equation (3.43) we see that the quantity r d−1 qr is a constant independent of the
radial distance. Note that effective permeability of a heterogeneous medium should
be independent of fluid properties. Thus, usually for simplicity, it is derived under
the assumption of an incompressible fluid. In this case r d−1 qr is just proportional
to the mass flux of the fluid. Using this constant we rewrite Darcy’s law in the
following form:

∂ r 1−d
Pp = − r d−1 qr . (3.44)
∂r D(r )S(r )
For an arbitrary d a direct spatial averaging of (3.44) will not necessarily provide
any distance-independent hydraulic diffusivity. To find parameters of a homoge-
neous medium replacing the heterogeneous one we could apply spatial averaging
over a length 2L of the equation above and compare such an averaged equation
with a correspondingly averaged equation for a homogeneous medium. The spatial
averaging of a quantity X (r ) around a point r x is given by the integral
r x +L
X (r x , L) = (2L)−1 X (r )dr. (3.45)
r x −L
Taking into account that the fluid-mass flux is constant we obtain from equa-
tion (3.44):

∂ r 1−d
Pp = − r d−1 qr . (3.46)
∂r D(r )S(r )
For a replacing homogeneous medium with the effective averaged parameters De f f

and Se f f this equation must have the following form:
1−d
∂ r
Pp = − r d−1 qr . (3.47)
∂r De f f Se f f
The product De f f Se f f must be equal to the ratio ke f f /ηe f f . Assuming a homo-
geneous fluid saturating the rock we obtain De f f Se f f = ke f f /η. In the original
heterogeneous medium, D(r )S(r ) = k(r )/η. Further, equations (3.46) and (3.47)
are equivalent under the following condition:

−1 1−d −1 r 1−d
(De f f Se f f ) = r . (3.48)
D(r )S(r )
From here we obtain several simple permeability-upscaling rules for steady-state
filtration in a heterogeneous medium. Assuming a homogeneous distribution of the
viscosity in a 1D medium (d = 1), we obtain a well-known rule:

−1 1
ke f f = . (3.49)
k(r )
Neglecting the elastic heterogeneity of the medium (especially in comparison with
its hydraulic permeability) we can apply exactly the same rule for the diffusivity:

−1 1
De f f = . (3.50)
D(r )
In 1D statistically homogeneous structures the spatial averaging is independent of
r x . The averaging stabilizes with increasing L. Thus, the effective diffusivity (and,
correspondingly, the permeability) indeed becomes a constant in the limit of large
averaging lengths L.
The situation becomes less convenient in 2D and 3D media, where the effective
parameters are explicit functions of the averaging length L and position r x . Indeed,
in 2D media we obtain

−1 r x + L −1 1
(De f f Se f f ) = 2L ln . (3.51)
rx − L r D(r )S(r )
In 3D media the analogous result is

−1 1
(De f f Se f f ) = (r x2 −L ) 2
2
. (3.52)
r D(r )S(r )
Corresponding averaging rules in a piecewise-homogeneous medium composed
of homogeneous layers of the thickness h = 2L/Nh (here Nh is the total num-
ber of layers) and different diffusivities Di (i = 1, 2, . . . , Nh is the number of a
corresponding layer) are following. In 1D media we have:
1
De−1 −1
f f = Nh i . (3.53)
Di
In 2D media we obtain
−1
rx + L 1 rx − L + i h
De−1 = ln i ln . (3.54)
ff
rx − L Di r x − L − h + i h
In 3D media the analogous result is
(r x2 − L 2 ) 1 1
De−1
ff = i . (3.55)
Nh Di (r x − L + i h)(r x − L − h + i h)
These results coincide with the rules derived in chapter 12 of Crank (1975).
However, in the case of a non-steady-state filtration these results are not nec-
essarily applicable. To consider such situations we must return back to equation
system (3.40) and (3.41). It is convenient to write it in a matrix form:
d−1
∂ r d−1 qr 0 −r d−1 S(r ) ∂t∂ r qr
= −1 1−d . (3.56)
∂r P p −(D(r )S(r )) r 0 Pp
The quantities r d−1 qr and Pp are continuous functions of r (even in the case of
layer interfaces; note again that we assume a homogeneous fluid). Let us consider
a sufficiently small d-dimensional spherical shell of thickness 2L δα , where
√
δα = Dt/π is a distance reciprocal to the attenuation coefficient of the slow
wave (2.105). Inside the 2L-averaging length, both the pore pressure and the fluid
flux will be nearly r -independent. Further, we are interested in filtration features
on distances r significantly larger than 2L. So r x can be identified with r . The
averaging on the scale 2L yields:
d−1
∂ r d−1 qr 0 −r d−1 S(r ) ∂t∂ r qr
= . (3.57)
∂r Pp −(D(r )S(r ))−1 r 1−d 0 Pp
The elements of the matrix in the right-hand side of this equation provide us with
averaging rules. They imply that averaging of the diffusivity coinciding with (3.48).
Therefore, features of a non-steady-state filtration on distances significantly larger
than δα will be defined by effective parameters corresponding to averaging rules
(3.49)–(3.50). These distances are of the order of several 10−1 parts of a slow-wave
length or larger. These averaging rules are independent of the dimensionality of the
medium. They correspond to a propagating slow plane wave in a local 1D hetero-
geneous structure. This is analogous to the eikonal approximation of the previous
section. This approximation implies an assumption of locally plane wavefronts too.
Note, however, that in contrast to the eikonal approximation, the radial scale of
heterogeneity we consider here is assumed to be significantly smaller than δα .
However, averaging rules indicated in (3.57) and correspondingly in (3.49)–
(3.50) do have a broader applicability. Let us assume that there exists a statistical
ensemble of various realizations of our radially heterogeneous model of the

medium. If the angular brackets denote the ensemble averaging (i.e. the statisti-
cal averaging) then its application to equation (3.56) will again provide equation
(3.57) for the ensemble-averaged field quantities r d−1 qr and Pp . Then, for the sta-
tistical ensemble averaging rules (3.49)–(3.50) become exact. In reality we have
one single realization of the medium only. However, in real media with ran-
dom small-scale 3D spatial fluctuations of physical properties and with randomly
distributed critical pore-pressure perturbations C, fitting smooth triggering-front
envelopes of microseismic clouds is similar to constructing averaged surfaces
of constant pressure (isobars). The ergodicity assumption (Rytov et al., 1989)
states then the equivalence of such a spatial averaging to the ensemble averaging.
Hummel and Shapiro (2012) considered several numerical examples of non-steady-
state filtration and demonstrated rather well the applicability of averaging rules
(3.49)–(3.50).
3.7 Back front of seismicity

If the injection stops at time t0 then induced earthquakes will gradually cease to
occur. For any time moment after time t0 a surface can be defined that describes
spatial positions of a maximum pore-pressure perturbation at this time. This surface
(it is a sphere in homogeneous isotropic rocks) separates the spatial domain, which
is still seismically active, from the spatial domain (around the injection point),
which is becoming seismically quiet. This surface was first described by Parotidis
et al. (2004) and termed the back front of induced seismicity (see Figure 3.14).
In simple situations this surface can be analytically described. Let us approx-
imate a fluid-injection borehole experiment by a point source of a constant
strength q, and the duration t0 , i.e. a boxcar function of duration t0 . The source
strength q (its physical unit is the watt) is related to the source term C f from
equations (2.219)–(2.221):
q = 4πC f D. (3.58)
Let PH (r, t) be a solution for the pore pressure during the injection time. Note that
this is a response of the medium to a Heaviside-source function. Because the sys-
tem under consideration is linear, then its response PB (r, t) to a boxcar-source
function will be the following difference of the two Heaviside-source-caused
responses:
PB (r, t) = PH (r, t) − PH (r, (t − t0 )). (3.59)
Figure 3.14 shows a pore-pressure distribution in a 2D medium for two different

points. One of them is close to the injection point. Another one is far from it. For
x 107 t0 t1
2
after
q [Pa]
1
before
0
x 105
r = 50 m
2
p [Pa]
0
x 104
6
r = 150 m
4
p [Pa]
0
0 0.5 1 1.5 2
t [s] x 105
Figure 3.14 Pore-pressure evolution before and after injection termination. The
back-front arrival at a given location corresponds to the maximum pore-pressure
perturbation achieved at this location. (Modified from Parotidis et al., 2004.)
small r , pore pressure rises immediately with the start of an injection and drops
promptly after t0 . For larger r , pore pressure begins to increase after the begin-
ning of the injection. This occurs until the time t1 , where the maximum pressure
value is reached. The time difference t1 − t0 increases with the distance r . For a
constant diffusivity D, t1 depends solely on r . Assuming that events may be trig-
gered for increasing pore-pressure values only, we expect no seismic activity after
the maximum-pressure arriving time t1 (r ). Therefore, a termination moment of the
seismic activity should correspond to the moment of the vanishing-time derivative
of the pore pressure. This time moment is a solution of the following equation:
∂
0= [PH (r, t) − PH (r, (t − t0 ))] . (3.60)
∂t
Note, however, that a time derivative of a Heaviside-function response of a linear

system is equal to its Green’s function (we denote it here as PG (r, t)). Thus, the
equation for the back-front surface becomes
0 = PG (r, t) − PG (r, (t − t0 )). (3.61)

The 1D, 2D and 3D Green’s functions for equation (3.1) in isotropic homoge-
neous media have the following form (see Carslaw and Jaeger, 1973, p. 262, and
Landau and Lifshitz, 1991, section 51.7), respectively:

q r2
PG1 (r, t) = exp − . (3.62)
(4π Dt)d/2 4Dt
Here d = 1, 2, 3 is the dimension of the space where the pressure diffusion occurs.
For example, in the normal 3D space it is equal to 3. In a 2D fault it is equal to
2. In a 1D hydraulic fracture (later we will address such a situation) it is equal
to 1. For given diffusivity D, and injection duration t0 , the solutions for t (r ) in
equation (3.61) give time moments of pore-pressure maximums at the correspond-
ing distance r . Writing this solution in a form of distance as function of time r (t)
yields the formulation of the radius Rb f ≡ r (t) of the back front:

t t
Rb f (t) = 2d Dt − 1 ln . (3.63)
t0 t − t0
Along with the triggering front, the back front is also a kinematic signature of
the pressure-diffusion-induced microseismicity. It is often observed on real data
(see Figure 3.15). In situations where the injection has produced a very moder-
ate or even zero impact on the permeability, the back front provides estimates of
hydraulic diffusivity consistent with those obtained from the triggering front and
approximately coinciding with the diffusivity of virgin rocks.
Fenton Hill Soultz 93

1200 2000
1500
800
r [m]
r [m]
1 1000
1
400 2
500 2
0 0
0 20 40 60 80 0 200 400
t [h] t [h]
Figure 3.15 Examples of r –t plots with triggering fronts (1) and back fronts (2)
of fluid-induced seismicity. The left-hand part and the right-hand part of the plot
correspond to the case studies of geothermal borehole-fluid injections at Fenton
Hill (an approximate estimate of the hydraulic diffusivity by the both curves is
0.14 m2 /s) and Soultz (an approximate estimate of the hydraulic diffusivity by
the both curves is 0.5 m2 /s), respectively. (Modified from Parotidis et al., 2004.)
3.8 Strength of pre-existing fractures

Besides hydraulic properties of rocks, the strength of pre-existing cracks is another
important parameter that influences fluid-induced microseismicity. This strength
corresponds to a value of the pore-pressure perturbation necessary to cause sliding
along a pre-stressed fracture, i.e. to trigger an earthquake. This is the previously
introduced criticality C of the medium. Thus, criticality is a pressure perturba-
tion that, according to the Mohr–Coulomb criterion (1.66), results in rock failure.
Variations of the criticality are caused by a combination of variations in the
cohesion, friction coefficient and the pre-injection stress field. Langenbruch and
Shapiro (2014) have recently shown that heterogeneities of elastic properties of
rocks strongly contribute to fluctuations of C. If C is high in a given location, we
consider this region to be stable. If C is low, we refer to this location as unstable,
with a higher probability that events are triggered due to pore-pressure changes.
In these terms, high criticality corresponds to high strength and low criticality
indicates low strength, respectively.
Rothert and Shapiro (2007) found a way to estimate the probability density
function f (C) of the medium criticality from microseismic data. Here, we pro-
vide a short review of their approach. In general, the probability density function
describes the frequency of a specific value of critical pore pressure in the corre-
sponding statistical ensemble. This ensemble is replaced by a single realization of
the medium under the ergodicity assumption.
3.8.1 Statistics of rock criticality

Let us assume the simplest possible distribution of the critical pore pressure, a
uniform probability density function f (C) for the criticality C(r) at any point with
position vector r; Cmin and Cmax will denote the minimum and maximum possible
criticality values, respectively. Then f (C) = 1/C, with C = Cmax −Cmin . The
statistical properties of C are assumed to be independent of position. Therefore,
C(r) is a statistically homogeneous random field. A criticality field randomly dis-
tributed within a number of cells Nc in three dimensions is shown in Figure 3.16
(two examples of 2D distributions of C are shown in Figures 3.6 and 3.7). The
gray scale corresponds to the criticality value at a given point. The probability
density function (PDF) is given by the number of cells in the medium asso-
ciated with a specific value of criticality normalized by the total number of
cells.
We consider a point-injection source located at the origin of a Cartesian
coordinate system. It causes pore-pressure perturbations Pp (r, t) in an infinite
homogeneous isotropic continuum with hydraulic diffusivity D. The pore-pressure
z [gp`s]
50
100
100
100
50 50
y [gp`s] 0 0
x [gp`s]
Figure 3.16 A synthetic distribution of the strength of pre-existing fractures (i.e.

criticality C) in a 3D rock model. (Modified from Rothert and Shapiro, 2007.)
perturbation evolves in space outside the source according to the diffusion equa-
tion (3.1). A fluid injection starts at time t = 0 with a constant injection rate. We
assume that for t < 0 the pore-pressure perturbation Pp is zero. Such a point source
of injection with a time dependence given by a step function can be described by
its duration t0 and its strength q defined in (2.219)–(2.221) and (3.58). We will
consider times less than t0 .
In the medium, the pore pressure Pp (r, t) rises monotonically as a function of
t ≤ t0 . The familiar solution of the diffusion equation (see equations (2.164) and
(2.196)) satisfies the given initial and boundary conditions in three dimensions:

q |r|
Pp (r, t) = erfc √ . (3.64)
4π D|r| 4Dt
An induced microseismicity can be then modeled as described above in previous
sections and shown in Figures 3.6–3.8.
To trigger a seismic event the following relationship between pore pressure and
criticality must be fulfilled:
Pp (re , t) = C(re ). (3.65)
In the case of a homogeneous and isotropic medium it means

q |re |
erfc √ = C(re ), (3.66)
4π D|re | 4Dt
where re are all points where events have been triggered at time t.
To reconstruct the probability density function of C, we have to compute a
number of events occurring in a spatio-temporal domain characterized by a crit-
icality between C and C + dC. For analysis of a given microseismicity cloud
using equation (3.66), the source strength q and hydraulic diffusivity D must be
known. For the case of numerical data, q and D are known from the model set-up.
For real injection experiments, the hydraulic diffusivity D can be obtained from
independent measurements like borehole tests or core studies. Alternatively, it can
be estimated from microseismicity according to the triggering front or back front
approaches described above. The source strength q can be estimated by using the
pressure at the borehole head (see equations (2.219)–(2.221) and (3.58)):
q = 4π p0 Da0 . (3.67)
A probability that a microseismic event occurs in a cell with criticality values
between, for example, C1 and C2 = C1 + dC (dC is small) is given by the
probability density function of criticality:

C1 + C2
WC1,2 = f · dC. (3.68)
2
On the other hand, this probability is equal to the fraction of cells with criticality
between C1 and C2 in the spatial domain, where until the observation time t the
pore pressure has already arrived in the range C1 ≤ Pp ≤ C2 . This spatial domain
is approximately a spherical volume with radius (r1 + r2 )/2 such that Pp (r1 , t) =
C1 and Pp (r2 , t) = C2 .
To count the number of events triggered in the time interval from 0 to t and
characterized by C1 ≤ C ≤ C2 , it is convenient to introduce a new coordinate
system (y, t). The variable y is defined as follows:
r
y = r0 ≡ √ , (3.69)
4Dt
where r is the distance from the injection point and t is the occurrence time of
a microseismic event. We can compute the criticality at a point with coordinates
(y, t) of a given event:
q
C= √ erfc(y). (3.70)
4π D 4Dt · y
Solving equation (3.70) for t we get
q2
t= · erfc2 (y). (3.71)
64π 2 D 3 y 2 C 2
Given the source strength q and hydraulic diffusivity D, this relationship between t
and y must be valid for each fluid-induced event characterized by the criticality C.
The criticality C is just a parameter of a curve t (y). Thus, the number of events with
the criticality C1 ≤ C ≤ C2 is just a number of events between the corresponding
curves t (y, C1 ) and t (y, C2 ).
Numerical data 14 349 events Criticality limits: 1000 – 40000 Pa

100
1
80
Normalized number of events

0.8
60
Time [h]
0.6
40
0.4
20 0.2
Reconstructed criticality
0 0
0 0.2 0.4 0.6 0.8 1 1.2
103 104
y [–]
C [Pa]
Figure 3.17 Left: t (y) curves (3.71) for a numerically simulated microseismicity
(dots) triggered using a criticality field with a uniform PDF. Right: a recon-
structed PDF of criticality (logarithmic sampling). (Modified from Rothert and
Shapiro, 2007.)
Therefore, the reconstruction of the criticality PDF can be carried out in the
following way. The y–t domain is subdivided into a number of sub-domains with
different criticalities (see the left-hand side of Figure 3.17). Note that equally
spaced criticality values result in non-equally spaced domains (y, t). The criti-
cality allocation can also be spaced logarithmically, resulting in a nearly equally
spaced (y, t) domain. The numbers of events located between sequential curves
ti (y) and ti+1 (y) given by criticalities Ci and Ci+1 , respectively, are calculated.
The best choice of the maximum-observation time tmax is usually the complete
injection time t0 . Usually, it provides the most complete statistics of events. At
the maximum time tmax values y, being solutions of the equations tmax = ti (y)
and tmax = ti+1 (y), correspond to spherical surfaces of the radii ri and ri+1 sat-
isfying the equations Pp (ri , tmax ) = Ci and Pp (ri+1 , tmax ) = Ci+1 . Thus, the
counted numbers of events represent the number of locations with criticalities
between Ci and Ci+1 in the spatial ball of an approximate radius (r1 + r2 )/2.
Therefore, this number of events must be normalized by the volume of such an
average ball and associated with the mean value of criticality (Ci + Ci+1 )/2.
Of course, such a normalization is an approximation. The smaller the difference
Ci+1 − Ci , the better the approximation. Additionally, if criticalities are spaced
non-uniformly, then the normalization procedure will, to some extent, also com-
pensate for this. Figure 3.17 gives an example of how this algorithm works on a
synthetic data set.
Often, hydraulic properties of rocks are anisotropic. In the case of a homoge-
neous anisotropic medium, the equation of pore-pressure diffusion (3.1) simplifies.
In the principal coordinate system of the diffusivity tensor we obtain equation

(3.13). In coordinate system (3.14) it turns into equation (3.15) describing the
relaxation of pore pressure in an isotropic medium with the scalar hydraulic dif-
fusivity Dii /3. The solution of the diffusion equation for the anisotropic case in
analogy to equation (3.64) becomes
−1
4π Dii 3/2 X el
Pp (r, t) = ( ) X el erfc √ , (3.72)
q 3 4t
where we have introduced the following notation:
x12 x2 x2
X el = + 2 + 3 . (3.73)
D11 D22 D33
We see that a hydraulic anisotropy influences the pore-pressure perturbation.
Also pore-pressure-induced clouds of microseismic events will show anisotropic
behavior. Microseismic clouds will tend to be ellipsoidal. Such ellipsoidal forms
are clearly observed especially in the triggering-time normalized space (see
Figure 3.11).
Therefore, to apply the criticality-reconstruction approach, we have to transform
a microseismic cloud obtained in an anisotropic medium into a cloud that occurs
in an effective isotropic medium. The coordinate system must be rotated into a
system parallel to the principal axes of the diffusivity tensor. After the rotation, all
coordinates must be additionally scaled as shown in equation (3.14).
Rothert and Shapiro (2007) reconstruct probability density functions of rock crit-
icality for the two already-presented real data sets obtained during experiments at
Hot Dry Rock (HDR) geothermal sites. Below we summarize and further interpret
their results.
3.8.2 Case study: Soultz-sous-Forêts

The Soultz-sous-Forêts Hot Dry Rock site is located in the Rhine Graben in eastern
France. During a geothermal stimulation experiment performed by SOCOMINE
(France) and CSMA (UK) in 1993, about 25 300 m3 of water were injected
into crystalline rocks over a depth interval of 2850–3400 m of borehole GPK1
(Dyer et al., 1994, Cornet, 2000). Flow logs indicate that major fluid loss (about
60%) occurred at a depth of approximately 2920 m (for details see Cornet
et al., 1997; Shapiro et al., 1999). During 400 h after the start of the injection,
about 9300 events were recorded and later located with a location error of approxi-
mately 20–80 m. The cloud of events is shown in Figure 3.4. Events recorded after
the stop of the injection (approx. 120) are not considered.
Shapiro et al. (2003) estimated a tensor of hydraulic diffusivity with an approxi-

mate relation of the principal components of 1:2.5:7.5. The largest-component axis
of the tensor of hydraulic diffusivity is close to vertical one. Its largest horizontal
component has nearly N/S direction. Its orientation is close to the orientation of
maximum horizontal stress (see Dyer et al., 1994; Cornet et al., 1997; Klee and
Rummel, 1993).
Owing to the fact that the injection-pressure function for Soultz can be rather
well approximated by a step function, equation (3.71) is approximately valid and
the criticality estimation approach can be applied. To estimate the PDF of the
criticality, Rothert and Shapiro (2007) proceeded as follows.
In equation (3.71), tmax = 400 h was used for the maximum time of pressure per-
turbation. They considered a pore-pressure perturbation along a cylindrical source
of length l = 100 m and radius Rc = 0.2 m. Surface pressure, i.e. well-head
pressure (which can be also taken as an approximation of the injection pressure),
was approximately constant at the level of 12 MPa (see figure 1 in Cornet, 2000).
To use the method of the previous section based on a simple solution for a spher-
ical injection cavity, the following far-field approximation was applied. The real
injection source was replaced by a surface-equivalent sphere. The radius of a
surface-equivalent sphere is then a0 = 3.2 m. According to equation (3.67) this
results in a point source of strength q ≈ 25 · 106 W. After scaling of the cloud into
one that would occur in an effective isotropic medium, the algorithm for criticality
reconstruction was applied. The maximum principal component of the hydraulic
diffusivity tensor of D =0.05 m2 /s was used. The normalized numbers of events
within the criticality shells was computed. The result is shown in Figure 3.18.
The reconstructed probability density function is shown on Figure 3.19 in loga-
rithmic and linear scales. The probability density function for criticalities lower
than 0.001 MPa is nearly vanishing. Then it increases, and for criticality values
above 3 MPa it steeply drops off.
Therefore, criticality for Soultz seems to be distributed in a broad range (three
orders of magnitude) between 0.001 MPa and approximately 3 MPa. The bounds of
the criticality limits seem to be sharp. The lower bound (Cmin ) is possibly defined
by the magnitude of tidal stresses, which are of the order of 0.001 MPa. These
stresses are permanently occurring and relaxing in the Earth. Thus, rocks in situ
seem to be accustomed to such perturbations and, thus, do not show significant
seismic reactions to such or smaller stress changes. The probability density func-
tion of criticality has an asymmetric shape. It does not seem to be Gaussian. It is
rather roughly uniform or a truncated Gaussian one.
The upper bound (Cmax ) is possibly defined by the injection conditions. Indeed,
as a result of replacing an open-hole-interval source by a surface-equivalent
spherical cavity this quite rough evaluation approach tends to underestimate the
Cmin=
350
100 Pa
300
250
Time [h]
200
150
100
Cmax =
50
20 MPa
0
–0.5 0 0.5 1 1.5 2 2.5 3 3.5
y [–]
Figure 3.18 A y–t plot for seismicity induced in the Soultz 1993 geothermal
experiment. (Modified from Rothert and Shapiro, 2007.)
Reconstructed criticality Reconstructed criticality

1 1
0.8 0.8
0.6 0.6
0.4 0.4
0.2 0.2
0 0
0 2 4 6 8 104 106
C [Pa] 6
x 10 C [Pa]
Figure 3.19 Reconstruction of the probability density function (PDF) of critical-

ity for the Soultz 1993 data set. The PDF is given in logarithmic and linear scales.
(Modified from Rothert and Shapiro, 2007.)
criticality. On the other hand it is clear that, even in the case of an exact pore-
pressure modeling, the maximum C estimates cannot exceed the injection pressure
of 12 MPa. To adjust these two values we must multiply the criticality range
obtained above by approximately 4. Then we will obtain values of the criticality
that are rather overestimated. This indicates that minimum pore pressures of 0.01
MPa are sufficient to seismically activate some pre-existing fractures at Soultz.
This value is in agreement with the estimates following from the case study KTB
considered at the beginning of this chapter. Necessary pressures for the majority
of seismic events are approximately between 0.01 MPa and 10 MPa. However, the
upper limit here is probably defined by the injection pressure. Thus, one cannot
exclude that Cmax can be even higher than 10 MPa.
3.8.3 Case study: Fenton Hill

Fenton Hill is located about 40 km west of Los Alamos on the west side of the
Rio Grande Graben in New Mexico, USA. During a massive hydraulic stimula-
tion experiment in 1983, about 21 600 m3 of water were injected at a depth of
approximately 3460 m for about 62 h (Fehler et al., 1998). Fluid injection took
place along an open-hole interval of approximately 20 m (House, 1987). Maxi-
mum bottom-hole pressure was approximately 48 MPa. During the experiment,
11 366 microseismic events were located with an accuracy better than 100 m
(House, 1987). Approximately 9350 events occurred during the time interval of
the injection and are shown in Figure 3.11.
Shapiro et al., (2003) estimated the tensor of hydraulic diffusivity with an
approximate relation of the principal components of 1:4:9. The largest axis of the
tensor points approximately in a NNW direction and dips with an angle of about
72◦ . The medium component points SSE with a dip angle of 18◦ . The smallest
component is about horizontal while pointing ENE. For the source strength q,
Rothert and Shapiro (2007) considered pore-pressure perturbation along a cylindri-
cal source of the length l = 20 m and the radius Rc = 0.1 m. Well-head injection
pressure, i.e. overpressure, was approximately constant at 14 MPa (House, 1987).
The radius of a surface-equivalent sphere is then a0 = 1 m. The diffusivity was
taken to be equal to 0.017 m2 /s. According to (3.67) this approximately results in
a source strength q of 35 ·106 W. After scaling in order to correct for anisotropy,
the algorithm described above was applied to the data. The results are shown in
Figures 3.20 and 3.21.
Comparable to the result for the Soultz data, the probability density function for
criticalities lower than 0.001 MPa is vanishing, then it increases until 0.01 MPa,
stays approximately constant until 0.1 MPa and then decreases towards 1 MPa.
Criticality over a broad range characterizes the strength of pre-existing fractures.
Again, the boundaries of the criticality limits seem to be sharp. The probability den-
sity function of criticality is again asymmetric (very roughly uniform or possibly
truncated Gaussian).
As in the previous case study, the values of C are possibly somewhat under-
estimated. We can additionally constrain these estimates. To match the highest
estimates of C to the injection pressure we must multiply all C values by
approximately 14. Then we will obtain values of the criticality that are rather
overestimated. Thus, as in the Soultz case study, this indicates that minimum
pore pressures of 0.01 MPa are sufficient to seismically activate some pre-existing
60
Cmin =
50 10 Pa
40
Time [h]
Cmax =
30 7 MPa
20
10
0
–0.5 0 0.5 1 1.5 2 2.5 3
y [–]
Figure 3.20 A y–t plot for seismicity induced in the Fenton Hill geothermal
experiment. (Modified from Rothert and Shapiro, 2007.)
Reconstructed criticality Reconstructed criticality

1
0.8 0.8
0.6 0.6
0.4 0.4
0.2 0.2
0 0
0 1 2 3 4 102 104 106
C [Pa] x 106 C [Pa]
Figure 3.21 Reconstruction of the criticality probability density function (PDF)

for the Fenton Hill data set. The PDF is given in logarithmic and linear scales.
fractures at Fenton Hill. Necessary pressures for the majority of events are approx-
imately between 0.01 MPa and 10 MPa. However, again one cannot exclude that
Cmax can be even higher than 10 MPa.
3.9 Spatial density of seismicity

Triggering front, back front and symmetry signatures of the fluid-induced
seismicity considered above characterize temporal evolution and geometry of
microseismic clouds. They describe kinematics of microseismic clouds. Already

in the previous section, where we studied statistics of critical pore pressure, we
worked implicitly with another type of seismicity characteristic that is related to
its spatial density. Real data usually indicate that the density (i.e. probability) of
seismic events is related to the magnitude of pore-pressure perturbations. Thus, we
call the density of events, event rates and event magnitudes the dynamic signature
of induced microseismicity.
We consider again a point source of a fluid injection in an infinite homogeneous
and isotropic porous continuum. Furthermore, we assume the same triggering
model of seismicity as in our previous discussions. The probability of an earth-
quake occurrence at a given point by a given time will be equal to W (C(r) ≤
pmax (t, r)), which is the probability of the critical pressure being smaller than
the maximum of the pore pressure Pp (t, r) achieved at this point by the given
time. If the pore-pressure perturbation caused by the fluid injection is a non-
decreasing function of time (which is approximately the case for step-function-like
borehole-injection pressures frequently used in reality) then this probability will be
equal to
Pp (t,r)
W = f (C)dC, (3.74)
0
where f (C) is the probability density function of the critical pressure. The pore
pressure Pp (t, r) is a solution of a diffusion equation describing pore-pressure
relaxation.
The simplest possible PDF of the critical pore pressure is a uniform one,
f = 1/C, where C is a normalizing constant. In the previous section we saw
that such a PDF can be accepted as a rough approximation of realistic criticality
distributions. In this case W = Pp (t, r)/C. Therefore, in such a simple model
the event probability is proportional to the pore-pressure perturbation.
Often realistic conditions of borehole fluid injections can be approximated by a
point source of pore-pressure perturbation (3.67) of a constant strength q switched
on at time 0. The corresponding solution of the diffusion equation then has the
form given by (3.64). For infinite observation time this equation reduces to (see
also Figure 3.22):
Pp (r, ∞) = q/(4π Dr ) = a0 p0 /r. (3.75)
The event probability is proportional to the volumetric density of microseismic

events. Therefore, a comparison of the volumetric event density with equa-
tion (3.64) can provide additional information on hydraulic properties of rocks. For
a synthetic cloud of microseismicity we count the number of events in concentric
spherical shells with the center at the injection point. Theoretically, the thickness
10–1
10–2
10–3
10–4
10–5
10–6
10 20 30 40 50 60
Distance [m]
Figure 3.22 Normalized pore-pressure perturbation (3.64) in a medium with a

given scalar diffusivity D due to a step-function point source of injection. For the
solid line D = 1 m2 /s and t = 10 000 s. For the dashed line D = 10 m2 /s and
t = 100 s. For the dashed-dotted line D = 1 m2 /s and t = 100 s.
of these shells must be differentially small. However, in order to ensure sufficient

statistics of events, the shells should not be too thin. In the case of real data, shell
thickness can be of the order of a spatial event-location error. Owing to normaliza-
tion, the central spherical volume defined by the internal spherical surface of the
first shell is not considered. By normalizing event numbers to the shell volumes we
obtain the event density. Then, we normalize all computed event densities by the
event density in the first spherical shell taken into account. This allows us to work
with non-dimensional quantities and to eliminate unknown proportionality factors.
To compare numerical results with predictions of equation (3.64), the analytical
function Pp (r, t) is also normalized by its value at the median radius of the first
spherical shell taken into consideration. The time in the error function is the total
period of the injection. Shapiro et al., (2005a,b) provide examples of corresponding
numerical simulations and show their agreement with (3.64).
Before such a comparison can be done for real data, the hydraulic anisotropy
of rocks must be taken into account. In order to compare a spatial distribution of
event density with analytical solution (3.64) we have to transform the microseismic
cloud. Analogously to our considerations in the previous section of this chapter,
this can be done by scaling an original event cloud along its principal axes and in
relation of the inverse square roots of the principal components of the hydraulic dif-
fusivity tensor (see equation (3.72)). Such a scaling procedure is a consequence of
the fact that, after such a scaling, the diffusion equation in an anisotropic medium
becomes equivalent to the diffusion equation in an isotropic medium. The hydraulic
Figure 3.23 A 2D projection of the seismicity cloud of the Soultz case study.
diffusivity in the resulting isotropic diffusion equation is equal to the arithmetic

average of the principal components of the diffusivity tensor. After the scaling, a
comparison between predictions of equation (3.64) and the event density is pos-
sible. It is completely analogous to the one in an isotropic medium (see Shapiro
et al., 2005a,b). Note that an inversion for a diffusivity tensor using a triggering
front is able to provide orientations and relations between the tensor’s principal
components with a rather high precision. However, their absolute values (i.e. the
magnitude of the arithmetic average of the principal components) are estimated
less exactly. This is a consequence of the heuristic nature of proportionality factors
in the triggering-front definition (3.6). This disadvantage can be compensated for
by using the analysis of the spatial density of events we describe here.
As case histories we consider again the two data sets already addressed above,
the microseismic clouds from Soultz and Fenton Hill. Figures 3.23 and 3.24 show
2D projections of the corresponding seismicity clouds.
As discussed above, the tensor of hydraulic diffusivity in Soultz is characterized
by a significant anisotropy with an approximate relation of principal components
in proportion 1:2.5:7.5. The average of the principal components of the hydraulic
diffusivity D can then be fitted to match the data. A best-fit curve is shown in
Figure 3.25. This fit provides an estimate of D = 0.03 m2 /s. We observe a very good
agreement of the event density with the theoretical curve given by equation (3.64).
Independent methods of estimating hydraulic diffusivity at the same location
yield rather similar values: the average principal component of the diffusivity ten-
sor can be computed from results given by Shapiro et al. (1999) to be equal to
D = 0.023 m2 /s. This estimate is also based on microseismicity but results from
using the triggering front in separate angular sectors around the injection source.
Figure 3.24 A 2D projection of the seismicity cloud of the Fenton Hill case study.
100
log10(normalized event number)
10–1
10–2
10–3
Observed curve
Theoretical curve, D = 0.03 m2/s
r1 = 100
10–4
0 200 400 600 800 1000
Distance [m]
Figure 3.25 Event density versus theory for the Soultz 1993 case study. (Modified
from Shapiro et al., 2005a.)
The estimate of the apparent hydraulic permeability from the same location based
on a borehole flow test and reported in Jung et al. (1996) also can be used to
compute the hydraulic diffusivity. This approximately yields D = 0.022 m2 /s.
In the previous sections we have shown estimates of D = 0.05 m2 /s obtained
from triggering fronts and back fronts fitted to the complete data set of the
Soultz experiment (see Figure 3.15). These approaches assume that the medium is
hydraulically isotropic. Therefore, they provide order-of-magnitude estimates only.
Uncertainties of the event-density-based approach to estimating average diffusivity
seems to be less, of the order of several tens percent (some numerical studies are
presented by Shapiro et al., 2005a,b).
Previously we have seen that, in the case of Fenton Hill, the tensor of diffusivity
is characterized by an anisotropy with an approximate relation of the principal
100
log10(normalized event number)

10–1
10–2
–3
10
10–4 Observed curve

2
Theoretical curve, D = 0.12 m /s
r1 = 90
0 200 400 600 800 1000
Distance [m]
Figure 3.26 Event density versus theory for the case study Fenton Hill. (Modified
from Shapiro et al., 2005a.)
components as 1:4:9. A least-squares-fit of analytical function 3.64 gives an

estimate of D = 0.12 m2 /s and is shown in Figure 3.26. Fitting of the triggering
front and the back front of seismicity (see Figure 3.15) provide a very close result:
D = 0.14 m2 /s.
Finally, Figure 3.26 shows some systematic deviations of the observed event
density from the theoretical curve. This is possibly an indication of some non-
linearity of the pore-pressure diffusion process in the Fenton Hill case study. The
next chapter considers seismicity triggering by non-linear pressure diffusion in
detail.
4
Seismicity induced by non-linear fluid–rock
interaction
We have already seen that the method of passive seismic monitoring has signifi-
cant potential for characterizing physical processes related to fluid stimulations of
rocks. One of its important modern applications is spatial mapping of hydraulic
fracturing. On the other hand, understanding spatio-temporal dynamics of micro-
seismic clouds contributes to reservoir characterization. It helps to monitor and to
describe hydraulic fractures.
We will start this chapter with a simple intuitive approach to the quantitative
interpretation of hydraulic-fracturing-induced seismicity. The approach is based on
a volume-balance model of the growth of long thin simple-geometry (nearly 1D)
tensile hydraulic fractures. Then we will introduce a case study from a gas shale
showing fracturing of a 3D rock volume. This motivates a more-general formula-
tion of non-linear fluid–rock interaction by hydraulic stimulations of reservoirs.
We show that linear pore-pressure relaxation and hydraulic fracturing are two
asymptotic end members of a set of non-linear diffusional phenomena responsi-
ble for seismicity triggering. We formulate a general non-linear diffusion equation
describing the pore-pressure evolution and taking into account a possibly strong
enhancement of the medium permeability. Both linear pore-pressure relaxation and
hydraulic fracturing can be obtained as special limiting cases of this equation.
From this formulation we derive an expression for the triggering front of fluid-
induced seismicity, which is valid in the general case of non-linear pore-pressure
diffusion. Our results are valid for an arbitrary spatial dimension of diffusion and
a power-law time dependence of the injection rate. They show that, the larger the
non-linearity of the fluid–rock interaction, the more strongly propagation of the
triggering front depends on the mass of the injected fluid. Further, we investigate
the nature of diffusivity estimates obtained from the triggering front of non-
linear diffusion-induced seismicity. Finally, we introduce a model of anisotropic
non-linear diffusion and show its application to the gas-shale data set mentioned
earlier.
164
4.1 Seismicity induced by hydraulic fracturing

Here, we show that evaluation of spatio-temporal dynamics of a cloud of induced
microseismic events can contribute to characterization of the hydraulic-fracturing
process. Again, r –t plots (plots of distances from the injection source to hypocen-
ters of seismic events versus their occurrence times) turn out to be a useful tool.
For example, r –t plots show signatures of fracture volume growth, of fracturing
fluid loss, as well as diffusion of the injection pressure into rocks and inside the
fracture.
4.1.1 Triggering front of hydraulic fracturing

During hydraulic fracturing a treatment fluid is injected through a perforated
domain of a borehole into a reservoir rock under a bottom-hole pressure larger
than the minimum principal stress σ3 . Often this stress is called the fracture closure
pressure, implying that uniaxial tensile strength from equation (1.79) is vanish-
ingly small. Usually it is expected that a tensile fracture (J = I , see Section 1.2.4)
will be opened (see Figure 4.1) so that its longest axis will be parallel to the max-
imum compressional tectonic stress σ1 and its shortest axis will be parallel to the
minimum compressional tectonic stress σ3 . Thus, depending on the tectonic stress
regime, the hydraulic fracture can be vertical (e.g. a normal stress regime) or hori-
zontal (e.g. a thrust stress regime). It can, of course, also be oblique in the case of
tilted tectonic stresses.
In order to understand the main features of induced seismicity by hydraulic frac-
turing we apply a simple model of the fracture-growth process resulting from a con-
sideration of the volume balance of an incompressible treatment fluid. The volume
balance requires the volume Q c (t) of the injected fluid to be equal to a sum of the
w
hr )
L(t
Figure 4.1 A geometrical sketch of a PKN hydraulic fracture.

166 Seismicity induced by non-linear fluid–rock interaction
stored-in-fracture fluid volume Q f (t) and of the lost-into-surrounding-formation

fluid volume Q l (t):
Q c (t) = Q f (t) + Q l (t). (4.1)
A straight planar height-fixed fracture (usually a vertical one – this is the case
for nearly all real-data examples given here) confined in a reservoir layer (see
Figure 4.1) is considered. This is the so-called PKN (Perkins–Kern–Nordgren)
model, which is very well known in the theory and practice of hydraulic fractur-
ing (see Economides and Nolte, 2003, and references therein; see also Peirce and
Detournay, 2008, for computational approaches to hydraulic fracture modeling).
Under these conditions, the half-length L of the fracture is approximately given
as a function of the injection time t by the following expression:
QI t
L(t) = √ , (4.2)
4h f C L 2t + 2h f w
where Q I is an average steady-state injection rate of the treatment fluid, h f is
an average fracture height and w is an average fracture width. The first term in
the denominator describes the fluid loss from the fracture into the rock. It is pro-
√
portional to t (this is a result of the approximation proposed by Carter, 1957)
and has a diffusion (filtration) character. The fluid-loss coefficient C L depends
on many factors, including the wall-building characteristics of the fracturing
fluid, the relative permeability of the reservoir rock to the fracturing filtrate, the
hydraulic diffusivity D of the medium, the injection pressure, etc. The fluid-loss
coefficient is an important reservoir engineering parameter and is a subject of
research. The second term, 2h f w, represents the contribution of the effective frac-
ture volume and depends mainly on geometry of the vertical cross-section of the
fracture.
In the case of hydraulic fracturing of a formation with a very low permeabil-
ity (e.g. gas shale or tight gas sandstones) the fracture body represents the main
permeable channel in the formation. A propagating fracture changes the effective
stress state in its vicinity and activates mainly slip events (observations of tensile,
implosive and CLVD events are also reported) in the critical fracture systems exist-
ing in surrounding rocks (Rutledge and Phillips, 2003). Thus, the fluid-induced
microseismicity is concentrated in a spatial domain close to the hydraulic frac-
ture (see also Warpinski et al., 2001). Figure 4.2 schematically shows such a
situation. Therefore, equation (4.2) can be considered as a 1D approximation
for the triggering front of microseismicity in the case of a penetrating hydraulic
fracture.
By the hydraulic fracturing of tight rocks (i.e. rocks with extremely low perme-
ability) this equation replaces the triggering front (3.6). In the limit of a long-time
Phase 1 Stage 2
300 cgu 22–10
250
200
150
y [m]
100
50
cgu 21–09
0
–50 cgu 21–10
–100
–300 –200 –100 0 100 200 300 400
x [m]
Figure 4.2 Schematic representation of a hydraulic fracture along with a cloud

of microseismic events at one of treatment stages in tight gas sandstones of the
Cotton Valley. (Modified from Shapiro, 2008, EAGE Publications bv.)
injection, the first (fluid-loss) term in the denominator of equation (4.2) becomes
dominant and, therefore, equation (4.2) becomes identical to equation (3.6):

L ≈ 4π Dap t, (4.3)
with apparent diffusivity Dap given by:
Q 2I
Dap = . (4.4)
128π h 2f C L2
Note, however, a completely different physical meaning of the apparent-diffusivity
coefficient Dap than the one of hydraulic diffusivity of rocks D from equation
(3.6). The latter is a material property, being the hydraulic diffusion coefficient.
On the other hand, Dap is nearly inversely proportional to the hydraulic diffusivity
of virgin rocks. It is a combination of geometrical factors and injection and rock
parameters.
If the volume of fluid loss is insignificant (e.g. this can be the case during the
initial phase of an injection), then the fracture length becomes close to a linear
function of time:
QI t
L(t) = . (4.5)
2h f w
Equations (4.2)–(4.5) give us a simple model of the growth process of a
hydraulic fracture during a fluid injection. They can be used as an approximation
of the spatial growth of a related microseismic cloud in the direction of fracture
penetration. In other words, they provide us with an approximate description of the
1D dynamics of the microseismic cloud during the injection phase.
4.1.2 Back front of hydraulic fracturing

What happens after the termination of the injection? To some extent the fracture can
still continue to penetrate. However, the dominant part of the seismicity induced
after an injection stop is located in the fractured domain (including the body of
hydraulic fracture usually filled with a proppant). We assume that the seismic-
ity is mainly triggered by the process of fluid-pressure relaxation in the fractured
domain. This hypothesis is also very close to the established understanding of
seismicity triggering by hydraulic fracturing (see Warpinski et al., 2001). The
process of pore-pressure relaxation can be approximately characterized by a sys-
tem of equations describing the dynamics of fluid-saturated porous elastic solids
(see Chapter 2). Such a description requires a more-precise treatment of fluid
properties. For instance, in contrast to volume-balance considerations, the fluid
compressibility must be taken into account.
We assume that the fractured domain is a thin planar infinitely long homoge-
neous strip of a porous elastic solid. The injection source is a thin straight line (a
perforated borehole) crossing this strip normally to its bottom and top. The pene-
tration depth of the treatment fluid and of the pore-pressure perturbation behind the
fracture walls is assumed to be very small in comparison with the fracture length.
Therefore, the geometry of the problem is effectively one dimensional. Let x be a
coordinate along the fracture length.
As an approximation we can further assume a fluid injection with a constant fluid
rate starting at time t = 0 at point x = 0 and terminating at time t = t0 . Assuming
that events may be triggered only for increasing pore-pressure values, this would
result in no later seismic activity at a point x, where Pp (x, t) has reached its max-
imum at the time t. Therefore, (x, t) coordinates where the fluid pressure Pp (x, t)
reaches its maximum give the distance xb f (t) (we assume a symmetric fracture
and consider the half-space with positive x) from the injection source to points
at which the induced microseismicity is terminated after the end of fluid injec-
tion. This distance determines the so-called back front of the induced seismicity
(Parotidis et al., 2004).
In the following discussion we will accept for the back front a rough approxima-
tion of a linear diffusion of the pore pressure in the volume of the stimulated rock.
In this case we obtain the following expression:

t t
xb f = 2Dt − 1 ln . (4.6)
t0 t − t0
This equation is identical to equation (3.63) with d = 1.

Equations (4.2)–(4.5) and (4.6) provide us with approximate bounds of
hydraulic-fracturing-induced microseismicity in the r –t domain.
4.1.3 Case study: Cotton Valley

We consider microseismic clouds (event locations courtesy of James Rutledge,
LANL) induced by hydraulic fracturing experiments in two boreholes of the
Carthage Cotton Valley gas field, East Texas, USA (see Figures 4.3 and 4.4). The
gas reservoir is a low-permeability interbedded sequence of sands and shales. We
will pay especial attention to the fracturing treatments called stages 2 and 3 of
treatment well 21-10 (see Figures 4.5, 4.6 and the upper part of Figure 4.4) and
described by Rutledge and Phillips (2003). Theses fracturing experiments are also
described as the treatments B and A in (Rutledge et al., 2004). There the hydraulic
fracturing was conducted at depths of 2757–2838 m and 2615–2696 m, respec-
tively. In both experiments the total perforation intervals were 24 m. About 990
and 650 microearthquakes, respectively, were located by using two borehole arrays
of 3-C geophones.
For all located microearthquakes of stage 3, Figure 4.7 (bottom) shows event
distances from corresponding nearest perforation points versus event occurrence
400
Stage 1 events
Stage 2 events
300 Stage 3 events
injection well
observation well
North [m]
200
100
−100
−400 −300 −200 −100 0 100 200 300 400 500
East [m]
400
Stage 1 events
Stage 2 events
300 Stage 3 events
injection well
observation well
North [m]
200
100
−100
0 100 200 300 400 500 600 700 800 900
East [m]
Figure 4.3 Map view of microseismicity induced by hydraulic fracturing at bore-

holes 21-09 (bottom) and 21-10 (top) in tight gas sandstones of the Cotton Valley.
(Modified from Dinske et al., 2010.)
−2500
Stage 1 events
Stage 2 events
−2600 Stage 3 events
perforated interval
Depth [m]
−2700
−2800
−2900
−3000
−400 −300 −200 −100 0 100 200 300 400 500
East [m]
−2500
Stage 1 events
Stage 2 events
−2600 Stage 3 events
perforated interval
Depth [m]
−2700
−2800
−2900
−3000
0 100 200 300 400 500 600 700 800 900
East [m]
Figure 4.4 Vertical plane projection of microseismicity induced by hydraulic

fracturing at boreholes 21-09 (bottom) and 21-10 (top) in tight gas sandstones of
the Cotton Valley. (Modified from Dinske et al., 2010.)
Stage 3
150
Seismic Events
100 Well 21–10
50
North [m]
–50
–100
–150
–200 –100 0 100 200 300 400
East [m]
Figure 4.5 Map view of gel-fracturing stage 3 of Cotton Valley tight gas
sandstone. (Modified from Dinske et al., 2010.)
Stage 2
100
Seismic Events
50 Well 21–10
North [m]
0
–50
–100
–300 –200 –100 0 100 200 300
East [m]
Figure 4.6 Map view of gel-fracturing stage 2 of Cotton Valley tight gas reservoir.
Note an intersection of the hydraulic fracture of stage 2 with a pre-existing natural
fracture between 30 and 100 m east. About 40% of induced earthquakes occurred
along this natural fracture. Note also the difference in the event-location quality
between Figure 4.2 and this figure (obtained after a precise event re-picking; data
courtesy of James Rutledge, LANL). (Modified from Dinske et al., 2010.)
Cotton Valley – Phase 1 stage 3

200 50
flow rate bh pressure
Flow rate [1/s]
Pressure [Mpa]
150 45
100 40
50 35
0 30
0 1 2 3 4 5 6 7
Time [h]
Cotton Valley – Phase 1 Stage 3

500
events
400 Dtf = 0.65 m2/s
Distance [m]
300
200
100 Dbf = 4 m2/s Dbf = 3.2 m2/s
0
0 1 2 3 4 5 6 7
Time [h]
Figure 4.7 Hydraulic-fracturing treatment data (top) and r –t plot of induced

microseismicity (bottom) for fracturing stage 3 of well 21-10 of the Carthage
Cotton Valley gas field. (Modified from Shapiro et al., 2006b.)
time. In the same time scale Figure 4.7 (top) shows the flow rate of the treatment
fluid along with the measured bottom-hole pressure.
A parabolic envelope given by equation (4.3) seems to describe well the upper
bound of the majority of microseismic events. One possible explanation is that dur-
ing a large part of the fluid-injection period the fracture growth is mainly controlled
by fluid-loss effects. Apparent hydraulic diffusivity Dap = 0.65 m2 /s characterizes
these effects and can be further used in equation (4.4) in order to estimate the
fluid-loss coefficient (C L ). Indeed, from this equation, we see that
QI
CL = . (4.7)
8h f 2π Dap
In the experiment under consideration the fluid rate during the injection intervals
can be approximated as Q I = 0.12 m3 /s. The height of the fracture can be assumed
to be approximately equal to the total length of the perforated borehole interval,
h f = 24 m. Thus, we obtain the following order-of-magnitude estimate. C L =
3×10−4 m/s1/2 . It is known that, typically, the fluid-loss coefficient is of the order of
2×10−5 m/s1/2 to 2×10−3 m/s1/2 (see, for example, Economides and Nolte (2003)).
Thus, we obtain a very realistic result.
The back front of the induced microseismicity is clearly seen after the stop of
the fluid injection. The last (third) injection phase was completed with a cross-link
gel and proppant. The back front (see equation (4.6)) provides a high diffusivity
estimate Db f ≈ 3.2 m2 /s. A possible interpretation is that this estimate mainly
represents the hydraulic diffusivity along the propped hydraulic fracture saturated
by the treatment fluid. Moreover, even after the end of the second injection phase
(approximately 1 h after the injection start) a back-front signature can be seen in
the r –t plot. An even higher diffusivity estimate, Db f ≈ 4 m2 /s, is explained by
the fact that during the first two phases of the injection no proppant but rather
a low-viscosity fluid was used for the treatment. Thus, the fracture must have a
somewhat larger permeability (before it has been completely closed) than in the
case of a proppant-filled fracture. Also the fluid viscosity influences the diffusivity
estimates.
Figure 4.8 shows a zoom of the first 1.5 h of the treatment. It indicates one
more important signature of the process of hydraulic fracturing. During the first
approximately 10 min of fracturing a quick, quasi-linear-with-time growth of the
microseismic cloud up to 150 m in length can be observed. We can try to inter-
pret these data by using equation (4.5), where the process of new fracture-volume
creation is assumed to be dominant. For example, we can attempt to estimate the
average width of the hydraulic fracture:
QI t
w= . (4.8)
2h f L(t)
East [m]
–150 –100 –50 0 50 100 150
50
North [m]
Well 21–10 0 Stage 3
–50
–100
Distance [m]
300
250
200
150
200 100
50
150 0
0 1 2 3 4 5 6 7
Time [h]
100
50
0
0 0.5 1 1.5 2
Figure 4.8 The r –t plot of the microseismicity induced during stage 3 with indi-
cated features corresponding to the first opening and following reopening of the
fracture. The inset in the upper left-hand corner of the figure shows a map view
of locations of microseismic events from the vicinity of the straight line shown
in the time range of first 10 min of the r –t plot. (Modified from Shapiro, 2008,
EAGE Publications bv.)
The injection rate Q I averaged over the first 10 min was about 0.08 m3 /s. The
average fracture height can again be approximated by the total perforation interval,
i.e. 24 m. Figure 4.8 shows that during the first 10 min of the injection the fracture
reached approximately 150 m half-length. Thus, the average fracture width w is
approximately 7 mm. It is known that in low- to medium-permeability formations
(and this is the case here – we work with a tight gas sand) the average hydraulic
fracture width is in the range of 3–10 mm (Economides and Nolte, 2003). Again,
we arrive at a very realistic result.
Interestingly, such quasi-linear growth of the microseismic cloud can be
observed in the first 10–20 min of each of the three injection periods (see Fig-
ure 4.8 at 0 h, at approximately 40 min and 3 h after the start of the treatment). We
interpret this as a signature of the fracture closing after each injection stop and its
reopening at the start of following injection phases.
Very similar features (see Figures 4.9 and 4.10) can be also seen in the data of
the hydraulic fracturing experiment corresponding to stage 2 of the gel treatment.
For the fluid-loss coefficient we obtain C L = 3.7 × 10−4 m/s1/2 and the fracture
average width is approximately of the order of 8 mm.
200 55
150 50
Pressure [MPa]
Flow rate [I/s]
100 45
50 40
0 35
0 1 2 3 4 5 6 7 8 9
Time [h]
400
events
300 2
Distance [m]
Dtf = 0.3 m /s
200
100
2
Dbf = 1.1 m /s Dbf = 0.9 m2/s
0
0 1 2 3 4 5 6 7 8 9
Time [h]
Figure 4.9 Hydraulic fracturing treatment data (top) and r –t plot of induced
microseismicity (bottom) for the fracturing stage 2 in the well 21-10 of Carthage
Cotton Valley gas field. (Modified from Shapiro et al., 2006b.)
East [m]
–100 –50 0 50 100 150 200

50
North [m]
Well 21–10 0 Stage 2
–50
300
Distance [m]
250
200
150
100
50
150 0
0 1 2 3 4 5 6 7 8 9
Time [h]
100
50
0
0 0.5 1 1.5 2
Figure 4.10 The r –t plot of the microseismicity induced during stage 2 with indi-
cated features corresponding to the first opening and following reopening of the
fracture. The inset in the upper left-hand corner of the figure shows a map view
of locations of microseismic events from the vicinity of the straight line shown
in the time range of first 10 min of the r –t plot. (Modified from Shapiro, 2008,
EAGE Publication bv.)
−200 −150 −100 −50 0 50 100 150 200 −200 −150 −100 −50 0 50 100 150 200
100 100
events (0−600s) injection well events (0−700s) injection well
50 50
North [m]
North [m]
0 0
−50 −50
−100 100 100 −100

events (2000−2600s) events (3200−3800s)
50 50
North [m]
0 0
−50 −50
100 −100 −100 100

events (11000−11600s) events (6300−7300s)
50 50
North [m]
North [m]
0 0
−50 −50
−100 100 −100

−200 −150 −100 −50 0 50 100 150 200 events (15900−16900s)
East [m]
50
North [m]
−50
−100
−200 −150 −100 −50 0 50 100 150 200
East [m]
Figure 4.11 Temporal evolution of the hydraulic fractures of stages 3 (on the
left) and 2 (on the right) of the Cotton Valley case study in terms of induced
microseismic events (the map view).
Figures 4.8 and 4.10 help us to identify time intervals where the process of
opening of hydraulic fractures was interrupted and even turned to their closing.
Figure 4.11 shows microseismic images of the hydraulic fractures of both stages at
the final moments of their multiple reopening. For the images, those events were
selected that were used before for fitting straight lines on Figures 4.7–4.10.
In order to analyze a long-term opening process we consider composite micro-
seismicity, which would correspond to a treatment of several hours’ duration with
an approximately constant injection rate. To do this, we eliminate time intervals
corresponding to interruptions in the fluid injections at stages 2 and 3. Figure 4.12
shows composite r –t plots obtained in this way. We can clearly observe a pre-
dominantly diffusional character of the fracture growth. Such a corrected r –t plot
can help to further improve estimates of the fluid-loss coefficient. Also, the back-
front signatures can be nicely identified. How can we now use these signatures to
characterize the reservoir?
500 150 500 150
Events Events
Flow Rate [1/sec]

400 Flow rate 120 400 Flow rate 120
Flow rate [I/s]

Distance [m]
Distance [m]
300 90 300 90
200 60 200 60
100 30 100 30
0 0 0 0
0 1 2 3 4 5 6 0 1 2 3 4 5 6
Time [h] Time [h]
500 500
Triggering front Triggering front
400 Back front 400 Back front
Distance [m]
Distance [m]
300 300
200 200
100 100
0 0
0 1 2 3 4 5 6 0 1 2 3 4 5 6
Time [h]
Time [h]
Figure 4.12 Composite r –t plots of stages 3 (to the left) and 2 (to the right) cor-
responding to several-hours-long fluid injection with an approximately constant
injection rate. (Modified from Shapiro, 2008, EAGE Publication bv.)
4.1.4 Estimating permeability of virgin reservoir rocks

We can attempt to further interpret the fluid-loss coefficient neglecting
near-fracture-surface effects. This means that we neglect the effect of a possible
filter cake on the pressure difference p between fracture and far-field reservoir. In
addition, we neglect the effect of the hydraulic-fracturing process on the pore space
and pre-existing natural fractures in the vicinity behind the walls of the fracture.
If all these effects are significant, then our estimate below can significantly devi-
ate from the property of a virgin reservoir. It corresponds instead to an effective
permeability.
As an approximation, we can use the following equation (see A9-13 of Econo-
mides and Nolte, 2003):

C L ≈ k ϕ ct /(π η) · p, (4.9)
where k and ϕ are the permeability and porosity of the reservoir, and ct and η are
the compressibility and viscosity of the reservoir fluid, respectively. Combining
this equation with equation (4.7) we obtain
Q 2I η
k≈ . (4.10)
128h 2f p 2 ϕct Dap
For example, we can use the following typical values for the Cotton Valley
tight gas reservoir conditions (a hydrocarbon gas at temperature of 120 ◦ C and
pressure of 28 MPa): ϕ = 0.1; ct = 3.5 · 10−8 Pa−1 ; η = 3 · 10−5 Pa · s. Assuming a

hydrostatic far-field reservoir pressure and taking for p an estimate of 15 MPa, we
obtain for the virgin reservoir permeability an estimate of the order of 10−5 darcy.
This is in a good agreement with values used for engineering simulations at Cotton
Valley.
4.1.5 Estimating permeability of hydraulic fractures

The hydraulic diffusivity of a fractured domain can be approximately written
as (see our equation (2.99) and its discussion for highly porous rocks; see also
Economides and Nolte, 2003 chapter 12):
D ≈ k f /(c f η f ϕ f ), (4.11)
where k f is a permeability of the fractured domain, ϕ f is its porosity, and c f and
η f are the compressibility and viscosity of the treatment fluid, respectively. If we
assume that the back-front diffusivity Db f is equal to the diffusivity of the fracture,
we will obtain:
k f ≈ Db f c f η f ϕ f . (4.12)
For example, we use Db f ≈ 3 m2 /s, c f ≈ 1/(2.25 · 109 Pa) (we assume the
compressibility of the treatment fluid is of the same order as the compressibility
of water), η f = 150 · 10−3 Pa · s (viscosity of the cross-link gel, see Rutledge and
Phillips, 2003) and ϕ = 0.3 (an estimate of the porosity of the fractured domain
including the propped fracture). This provides us with an estimate of the effective
permeability k f of the hydraulic fracture of 70 darcy, which is a realistic value.
4.1.6 Case study: Barnett Shale

From the case studies of microseismicity we have considered so far we can
conclude that linear pore-pressure relaxation and hydraulic fracturing are two
asymptotic end members of a set of phenomena responsible for seismicity trig-
gering. Later, we will show that both types of seismicity-inducing processes (on
one hand a linear pore-pressure diffusion and, on the other, an opening of a new
volume in rocks) are limiting cases of a non-linear pressure-diffusion process lead-
ing to an enhancing of the medium permeability. Such a non-linear process seems
to be responsible for microseismicity triggering in shale. In particular, microseis-
micity of Barnett Shale shows signatures of non-linear-diffusion triggering with
an extremely strong hydraulic permeability enhancement in a 3D stimulated vol-
ume of a reservoir. This indicates a process of a volumetric hydraulic fracturing via
reopening of compliant pre-existing cracks and joints.
Features of hydraulic fracturing considered here are defined by specific
properties of a particular Barnett Shale gas reservoir (Fisher et al., 2002, 2004;
Maxwell et al., 2006). Features of the tectonic stress there (which still remains
to be better understood) are also significant. The Barnett Shale is a marine shelf
deposit of the Mississippian age. It is an organic-rich black shale with extremely
low permeability on the order of 0.1–0.5 microdarcy. Shapiro and Dinske (2009a)
have analyzed a microseismic data set from Barnett Shale (courtesy of Pinnacle
Technology). The data were obtained by hydraulic fracturing from a vertical bore-
hole in the Lower Barnett Shale of the Fort Worth Basin. These data demonstrate a
typical hydraulic fracture fairway network from a vertical well in the core area of
Barnett (Fisher et al., 2002).
One could propose the following scenario. We assume that the virgin reservoir
rock is extremely impermeable. This leads to a vanishing fluid loss from opened
fractures. We also assume that the treatment fluid is incompressible and that dur-
ing its injection it deforms and opens weak compliant pre-existing fractures in a
limited volume of the rock. We assume that the fractures can be opened if the pore
pressure exceeds a given critical value. As soon as a fracture has been opened the
permeability of the rock strongly increases, and the fluid can be further transported
to open more fractures. Note that fracture opening (i.e. fracture width) is a function
of the pore-pressure perturbation. However, in the case of extremely increasing per-
meability, one can assume that from the injection source up to the filtration front
(where Pp = 0) radial variations of the bulk porosity filled by the treatment fluid
are weak only. Then equation (4.5) should be replaced by
r (t) = A(Q I t/φ)1/3 . (4.13)
Here r (t) is a growing size of the fractured domain, φ is its porosity filled by the
treatment fluid and A is a dimensionless geometric factor equal, for example, to 1
or to (3/(4π))1/3 in the case of cubic or spherical fractured domain, respectively.
Therefore an r –t plot will show a t 1/3 parabolic envelope of corresponding
microseismic clouds. Exactly this type of behavior is demonstrated by the Bar-
nett Shale data set in Figures 4.13–4.15. Figure 4.13 shows the microseismic cloud
induced by the hydraulic fracturing. The spatial evolution of microseismicity with
injection time is also seen in the figure. Figures 4.14 and 4.15 show the injection
pressure, the fluid rate and the r –t plot for microseismic events. One can see that
the cubic-root type of the triggering front describes the data significantly better
than a square-root parabola.
The cubic-root parabola of Figure 4.15 and equation (4.13) permits us to esti-
mate the porosity opened and connected by the treatment fluid. An approximate
estimate of this additional (volume-averaged) opened porosity is of the order of
0.01%.
Therefore, in this particular case study, microseismic features of hydraulic frac-
turing correspond to a non-linear pressure diffusion in a medium with permeability
500 –2300
897 events
injection well
–2350
Depth [m]
–2400
–2450
0 897 events
perforation interval
–2500
North [m]
–200 0 200 400 600 800

East [m]
–2300
–500 –2350
Depth [m]
–2400
–2450
–1000 –2500
–200 0 200 400 600 800 –1000 –500 0 500
East [m] North [m]
Depth [m] Depth [m] Depth [m]
–2300 –2300 –2300

–2400 –2400 –2400
–200 0 200 400 600 800 –200 0 200 400 600 800 –200 0 200 400 600 800
–2300 –2300 –2300
–2400 –2400 –2400
–200 0 200 400 600 800 –200 0 200 400 600 800 –200 0 200 400 600 800
–2300 –2300 –2300
–2400 –2400 –2400
–200 0 200 400 600 800 –200 0 200 400 600 800 –200 0 200 400 600 800
East [m] East [m] East [m]
Figure 4.13 Hydraulic-fracturing-induced microseismicity in Barnett Shale (data

courtesy of Shawn Maxwell, Pinnacle Technologies). Top: map and vertical-plane
projections of the microseismic cloud. Bottom: spatial growth of the microseismic
cloud with time on a vertical-plane projection (note that the vertical and horizontal
scales are equal here; from Shapiro and Dinske, 2009a).
200 50
injection rate
pressure
150 45
Flow Rate [l/sec]
Pressure [MPa]
100 40
50 35
0 30
0 1 2 3 4 5 6 7
Time [h]
Figure 4.14 Borehole pressure (measured at the injection depth) and fluid flow
rate for the Barnett Shale data set shown in Figure 4.13. (Modified from Shapiro
and Dinske, 2009b.)
1000
800
Distance [m]
600
400
200
0
0 1 2 3 4 5 6 7 8
Time [h]
1000
800
Distance [m]
600
400
200
0
0 1 2 3 4 5 6 7 8
Time [h]
Figure 4.15 The r –t plot of induced microseismic events for the Barnett Shale
data set shown in Figure 4.13. Different envelopes are shown. Top: a linear-
diffusion-type approximation of the triggering (t 1/2 ); the dashed line gives a
possible indication of the back front. Bottom: a cubic-root parabola (t 1/3 ) better
matching the data. (Modified from Shapiro and Dinske, 2009b.)
that is strongly enhanced by the fluid injection. It seems that the volumetric tensile
opening of pre-existing compliant fractures embedded into extremely imperme-
able matrix is a dominant mechanism controlling the dynamics of the induced
microseismicity here. This process can be denoted as a 3D volumetric hydraulic
fracturing. The r –t plot shows a characteristic cubic-root parabolic behavior.
4.2 Seismicity induced by non-linear pressure diffusion

Above we have seen that spatio-temporal features of pressure-diffusion-induced
seismicity can be found in a very natural way from the triggering front concept.
In a homogeneous and isotropic medium the triggering front can be defined
corresponding to equation (3.6) in the limit of linear fluid–rock interaction. This
approximation might be sometimes suitable for Enhanced Geothermal Systems
Cooper Basin
80
70
Pressure [MPa]
60
50
40
30
20
10
0
0 1 2 3 4 5 6 7 8 9 10
Time [Days]
Flow Rate [l/sec] 60
50
40
30
20
10
0
0 1 2 3 4 5 6 7 8 9 10
Time [Days]
2000
Distance [m]
1500
1000
500
0
0 1 2 3 4 5 6 7 8 9 10
Time [Days]
Figure 4.16 Injection pressure, flow rate (on the two upper plots) and an r –t
plot of the corresponding fluid-injection-induced microseismicity at a geother-
mal borehole in the Cooper Basin, Australia. The data are courtesy of H. Kaieda.
(After Shapiro and Dinske, 2009b.)
(EGS). In the previous chapter we considered the case studies Soultz and Fen-
ton Hill in this approximation. Figures 4.16 and 4.17 show two more case studies
of EGS with the triggering front constructed in the linear-diffusion approximation.
Another extreme corresponding to a strong non-linear fluid–rock interaction is
hydraulic fracturing. Propagation of a hydraulic fracture is accompanied by open-
ing new fracture volumes, fracturing fluid loss and its infiltration into reservoir
rocks as well as diffusion of the injection pressure into the pore space of sur-
rounding rocks and inside the hydraulic fracture. In order to understand the main
features of the induced seismicity by such an operation, in the previous sections we
applied a simple approximation of the process of the fracture growth. We used a
volume-balance consideration for a single straight planar (usually vertical) fracture
confined in a reservoir layer. Half-length L of the fracture is approximately given
as a function of injection time t by expression (4.2). The fracture induces seismic
events in its vicinity. Thus, equation (4.2) is an approximation of the triggering
front in the case of a penetrating hydraulic fracture.
During the initial phase of hydraulic-fracture growth the contribution of the frac-
ture volume in the fluid volume balance is often dominant. This can lead to a nearly
Basel DHM
80 60
DH Pressure [MPa]
dh pressure
70 50
Pressure [MPa]
pressure
60 40
50 30
40 20
30 10
20 0
0 1 2 3 4 5 6
Time [Days]
4000
Flow Rate [l/min]
3000
2000
1000
0
0 1 2 3 4 5 6
Time [Days]
800
Distance [m]
600
400
200
0
0 1 2 3 4 5 6
Time [Days]
Figure 4.17 Injection pressure, flow rate (on the top two plots) and an r –t plot of
the corresponding fluid-injection induced microseismicity at a geothermal bore-
hole in Basel region of Switzerland. The data are courtesy of U. Schanz and M.
O. Häring. (After Shapiro and Dinske, 2009b.)
linear expansion with time of the triggering front (see equation (4.5)). This is an
example of 1D hydraulic fracturing. If the injection pressure drops, the fracture
will close. A new injection of the treatment fluid leads to reopening of the fracture
and, thus, to a repeated 1D hydraulic-fracture growth, i.e. repeated nearly linear (as
function of time) propagation of the triggering front. Because of a pressure depen-
dence of the fracture width and possible deviations of the fracture geometry from a
simple nearly rectangular planar shape, deviations from a linear-with-time fracture
growth can be observed. A long-term fluid injection leads to domination of diffu-
sional fluid-loss processes. The growth of the fracture slows down and becomes
√
approximately proportional to t. Figure 4.18 shows an example of data obtained
during hydraulic fracturing.
A process closely related to classical 1D hydraulic fracturing is the volumetric
(i.e. 3D) hydraulic fracturing considered in the previous section. It involves a nearly
incompressible treatment fluid, which during its injection deforms and opens weak
compliant pre-existing fractures in a limited rock volume (see Figure 4.13). An
r –t plot will show a t 1/3 parabolic envelope of corresponding microseismic clouds.
This type of behavior is shown by the Barnett Shale data set we discussed in the
previous section (see Figure 4.15).
Cotton Valley
200 50
Flow Rate [l/sec]
Pressure [MPa]
150 45
100 40
50 35
0 30
0 0.5 1 1.5 2 2.5 3
Time [h]
400
300
Distance [m]
200
100
0
0 0.5 1 1.5 2 2.5 3
Time [h]
Figure 4.18 Hydraulic fracturing induced microseismicity at the Carthage Cot-

ton Valley gas field. Top: borehole pressure (measured at the injection domain)
and fluid flow rate. Bottom: r –t plot of induced microseismic events. The dashed
line indicates approximately linear-with-time triggering front propagation corre-
sponding to equation (4.5). This figure reproduces the data set of Figure 4.9 in the
time period 4.5–7 h. (After Shapiro and Dinske, 2009b.)
Therefore, linear pore-pressure diffusion and non-linear fluid–rock interac-

tion can show different signatures of seismicity triggering fronts. Shapiro and
Dinske (2009a,b) proposed an approach unifying linear pressure diffusion and vol-
umetric hydraulic fracturing, i.e. equations (3.6) and (4.13). Following them we
derive a rather general non-linear diffusion equation describing the pore-pressure
relaxation. This equation takes into account a possibility of strong enhancement of
the medium permeability. Both asymptotic situations described above can then be
obtained in an arbitrary spatial dimension. In the next chapter we will further use
this formalism for a description of magnitude probabilities of induced events.
4.2.1 Non-linear diffusion and triggering fronts

Here we show that all types of scaling of microseismicity growth we reviewed
in the previous section are particular cases of more-general non-linear diffusion
dynamics. Under rather general conditions, including a possible strongly non-linear
interaction of an injected fluid with a rock, the pore-pressure relaxation can be
approximately described by a system of the two following differential equations.
The first one is the continuity equation expressing the fluid-mass conservation
(see equation (2.65)):
∂φρ f
= −∇qρ f , (4.14)
∂t
where ρ f is the density of a pore fluid, q is its filtration velocity and φ is the
rock porosity. Under realistic conditions (by neglecting irrotational skeleton defor-
mations and stress dependences of elastic properties) the time-dependent part of
quantity φρ f is proportional to the pore-pressure perturbation Pp (see equation
(2.113)), and can be substituted by ρ0 Pp S, with ρ0 being a reference fluid density.
We recall that S is a storage coefficient depending on the bulk modulus of pore
fluid K f , the bulk moduli of grain material K gr and of drained-rock skeleton K dr ,
the P-wave modulus of drained skeleton λdr + 2μdr and the Biot–Willis coefficient
α = 1 − K dr /K gr (see also equation (2.100)).
The second equation is the Darcy law, expressing a balance between the viscous
friction force and the pore-pressure perturbation:
k
q = − ∇ Pp , (4.15)
η
where k is the tensor of hydraulic permeability of the rock and η is the dynamic
viscosity of the pore fluid. In the case of non-linear fluid–rock interaction, the
permeability can be a strongly pressure-dependent quantity (see Section 2.9.7).
For simplicity we will neglect anisotropy and heterogeneity of rocks (a model of
an anisotropic non-linear diffusion will be considered later in this chapter). We
assume a spherical pore-pressure source switched on at time t = 0. Thus, we con-
sider a spherically symmetric problem in a d-dimensional space and combine the
two above equations in a corresponding d-dimensional spherical coordinate system
with the origin at the injection point (see also equation (3.42)):
∂r d−1 Pp ∂ ∂
= D(Pp )r d−1 Pp , (4.16)
∂t ∂r ∂r
where r is the radial distance from the injection point. We introduce a pressure-
perturbation-dependent hydraulic diffusivity:
k(Pp )ρ f (Pp )
D(Pp ) = . (4.17)
Sηρ0
The non-linear diffusion equation (4.16) must be completed by the initial condition
of zero pore-pressure perturbation before the injection starts (i.e. Pp = 0 for t < 0)
and by the two following boundary conditions. One of them gives the mass rate m i
of the fluid injection at the surface of an effective injection cavity of radius a0 :
k(Pp )ρ f (Pp ) ∂
m i (t) = −Ad a0d−1 Pp |r =a0 , (4.18)
η ∂r
where Ad will take one of the values 4π, 2π h, 2Ar , if the dimension d of the space
is 3, 2 and 1, respectively. In particular, h denotes the height of a hypothetical
homogeneous plane layer, where a cylindrically symmetric filtration takes place

(i.e. d = 2). Ar denotes the cross-sectional area of an infinite straight rod in the
case of a 1D filtration.
The second boundary condition states that a pore-pressure perturbation, along
with its spatial derivatives, is vanishing at infinity faster than 1/r d−1 .
Integrating equation (4.16) over distance r gives:
∞
∂ ∂
r d−1 Pp dr = D(Pp )r d−1 Pp |r =a0 . (4.19)
∂t a0 ∂r
Integrating this over time and taking into account the boundary condition (4.18) at
r = a0 gives:
∞ t
m i (t) m c (t) 1 t
Ad r d−1 Pp dr = dt = ≈ Q i (t)dt. (4.20)
a0 0 ρ0 S ρ0 S S 0
Here m c (t) denotes the cumulative mass of the fluid injected until time t. In the last
approximate term we have neglected pressure dependency of the fluid density and
we have introduced the volumetric rate of the fluid injection, Q i (t).
In the following we assume an injection source with a power-law injection rate
Q i (t) = S(i + 1)Ad Q 0 t i . For example, in the case of a constant-rate injection,
i = 0 and Q i = S Ad Q 0 . The dimensional constant Q 0 determines the strength of
the injection source (of course, along with the exponent i).
Further, we assume a power-law dependence of the diffusivity on pore-pressure
perturbations (see also equations (2.283)–(2.285)):
D = (n + 1)D0 Ppn . (4.21)
If n is large, the hydraulic diffusivity (and permeability, respectively) will depend

strongly on the pressure. If n = 0, the pressure relaxation will be described by
a linear diffusion equation. The exponent n can be considered as a measure (an
index) of non-linearity of the fluid–rock interaction.
Finally, we assume that the radius a0 of the injection cavity is vanishingly small
(compared to distances under consideration).
Under these assumptions, equations (4.16) and (4.20) take the following forms,
respectively:
∂r d−1 Pp ∂ ∂
= D0 r d−1 Ppn+1 , (4.22)
∂t ∂r ∂r
∞
r d−1 Pp (t, r )dr = Q 0 t i+1 , (4.23)
0
where the proportionality constants D0 and Q 0 are defined by properties of the

medium and of the injection source.
To understand features of the solution of the problem stated by equations

(4.22)–(4.23) we will apply the -theorem of the dimensional analysis (Baren-
blatt, 1996). The two equations above show that the pressure perturbation depends
on the following quantities: r , t, D0 and Q 0 . They have the following physical
dimensions:
L2 P Ld
[r ] = L , [t] = T, [D0 ] = , [Q 0 ] = , (4.24)
T Pn T i+1
where, L, T and P denote physical dimensions of length, time and pressure,
respectively. It is clear that three of the quantities r , t, D0 and Q 0 , arbitrarily cho-
sen, have independent physical dimensions. The physical dimension of one of them
(for example r ) can be expressed in terms of three others. In other words, only one
dimensionless combination, θ, can be constructed:
r
θ= n n(i+1)+1 1/(dn+2) . (4.25)
(D0 Q 0 t )
The following combination of the quantities t, D0 and Q 0 has the dimension of
pressure:
1/(dn+2)
Q 20
. (4.26)
D0d t (d−2i−2)
The -theorem of the dimensional analysis states then that the pressure must have
the following form:
1/(dn+2)
Q 20
Pp = (θ), (4.27)
D0d t (d−2i−2)
where is a dimensionless function, which must be found by solving the problem
formulated in (4.22)–(4.23) and the initial condition Pp = 0 for t < 0.
We see that the spatial distribution of Pp is completely defined by the dimen-
sionless parameter θ. If θ is large enough (i.e. small times and large distances from
the source) we expect an insignificant pressure increase. If θ is small (large times
and small distances) a strong change of pressure must occur. Thus, a constant value
of θ denotes a front of changing pressure. From an analysis of similar non-linear
equations (for example, analogous equations for very intensive thermal waves were
considered in Barenblatt, 1996, pp. 76–79, for the case d = 3) it is known that a
sharp front separating a zero-perturbation domain from non-vanishing perturbation
values can occur (see also our discussion at the end of this section and Figure 4.19).
Thus, a constant value of θ defines the triggering front Rt of induced seismicity.
From equation (4.25) we obtain a general result:
Rt ∝ (D0 Q n0 t n(i+1)+1 )1/(dn+2) . (4.28)

Let us first assume that the diffusivity is pressure independent. Then n = 0 and
√
Rt ∝ Dt. (4.29)
This corresponds to equation (3.6). The triggering of seismic events is controlled

by a linear pore-pressure relaxation (i.e. a linear diffusion process). Let us fur-
ther assume that n is large (this corresponds to an extremely non-linear diffusion
process). Then
Rt ∝ (Q 0 t (i+1) )1/d . (4.30)
In this case the triggering front is completely defined by the volume of the injected
fluid. In other words, equation (4.30) is an expression of the volume balance for an
incompressible fluid. The opening of a new pore (or fracture) volume occurs in the
spatial domain between injection source and triggering front. This volume is equal
to the volume of the injected fluid. For example, in the case of a hydraulic fracture
growth with d = 1 and a constant injection rate, i = 0,
Rt ∝ Q 0 t. (4.31)
This is in perfect agreement with equation (4.5), which describes the linear-with-
time opening of a single hydraulic fracture. The dashed line in Figure 4.18 indicates
such a process.
In the case of d = 3 and a constant injection rate, i = 0, we obtain:
Rt ∝ (Q 0 t)1/3 . (4.32)
This situation corresponds to equation (4.13) and Figure 4.15.

A special case of equation (4.22) with n = 1 corresponds to a classical equa-
tion describing the water head in an unsaturated medium. This equation was first
proposed and solved by the French scientist J. Boussinesq (Barenblatt et al., 1990;
Barenblatt, 1996). The Boussinesq equation with d = 2 and various values of i
corresponds to unsaturated flow to or from a borehole in a reservoir layer. The
parameterization n = 1, d = 1, and i = (3αi − 1)/2 corresponds to the initial con-
dition problem of the Boussinesq equation formulated by Barenblatt et al. (1990).
Particularly, the case of i = −1 (i.e. αi = −1/3) corresponds to a constant (i.e.
time-independent) amount of fluid instantaneously injected into the medium. This
yields the classical law Rt ∝ (D0 Q 0 t)1/3 of propagation of the saturation head in a
1D medium. In the case of i = −1 and a very strong non-linear diffusion (n 1),
equation (4.28) yields:
1/d
Rt ∝ Q 0 t 1/(dn+2) . (4.33)
Another form of the triggering front can be obtained by substituting the

cumulative mass of the injected fluid (equation (4.20)) into equation (4.28):
n
m c (t)
Rt ∝ dn+2
D0 t . (4.34)
Sρ0 Ad
Again we see that in the case of linear diffusion (i.e. n = 0) the injected mass
does not influence propagation of the triggering front. However, the stronger non-
linearity, the stronger is the effect of the injected fluid mass. In the limit of large n
we obtain:
m c (t)
Rt ∝ d
. (4.35)
Sρ0 Ad
Therefore, we conclude that, depending on parameter n, equation (4.22) describes

a broad range of phenomena: from a linear pore-pressure relaxation to hydraulic
fracturing.
Finally, we will attempt to analyze a full solution of problems (4.22)–(4.23).
Substituting Pp given by equation (4.27) into equations (4.22) and (4.23) we
obtain:
∂ 2 n+1 (d − 1) ∂n+1 n(i + 1) + 1 ∂ d − 2(i + 1)
+ + θ + = 0, (4.36)
∂θ 2 θ ∂θ dn + 2 ∂θ dn + 2
∞
θ d−1 (θ)dθ = 1. (4.37)
0
This equation system must be completed by the condition (∞) = 0.

For the set of constants d = 3, i = −1 (corresponding in our case to an
instantaneous injection of a finite volume of fluid into a 3D poroelastic medium),
equations (4.36)–(4.37) reduce exactly to the equation system (2.44)–(2.45) of
Barenblatt (1996). Thus, for our somewhat more-general problem, we try to find
a solution in a binomial form similar to those of Barenblatt (his equations (2.46)–
(2.50)). Using a direct trial we obtain that, under the condition i = −1, system
(4.36)–(4.37) has the following solution:
= J (θt2 − θ 2 )1/n , for θ < θt , and = 0, for θ > θt , (4.38)
where J = [n/(2(n + 1)(dn + 2))]1/n , and θt is another constant that can be found
from substituting solution (4.38) and consequent integrating in (4.37). For example,
the case of d = 1, n = 1, i = −1 corresponds to the Boussinesq equation analyzed
by Barenblatt et al. (1990). For this case they obtained θt2 = 8.
In this example the condition θ = θt defines a pressure front and, respectively,
a seismicity triggering front (exactly in the sense of equation (4.28)). From (4.38)
1
0.8
0.6
0.4
0.2
0
0 0.2 0.4 0.6 0.8 1
x
n = 1/2 n=1 n=2 n=3
Figure 4.19 Pressure (represented by the vertical axis and normalized to its
maximal value) as function of distance x (normalized to the triggering front)
corresponding to equation (4.39) for different values of n.
the following scaling can be obtained for the pressure on distances smaller than the
triggering front size (see Figure 4.19):
1/n
r2
p(t, r ) = p(t, 0) 1 − 2 . (4.39)
Rt
The pressure becomes vanishing small for larger distances. Moreover, we see
that, in the case of finite n ≥ 1, there exist a pressure front that is character-
ized by a discontinuous spacial derivative of the pressure. The triggering front of
seismicity can be directly associated with this pressure front. The pressure pro-
file has an abrupt character at the front in the case of a significant non-linearity.
Simultaneously, the pressure distribution behind the front becomes closer to a
rectangular one. We conclude that the stronger the non-linearity of the fluid–
rock interaction, the more uniform the pressure distribution behind the triggering
front.
4.2.2 Triggering fronts and diffusivity estimates

In this section we follow Hummel and Shapiro (2012) to investigate what kind of
diffusivity estimates will be obtained from fitting triggering fronts on r –t plots in
the case of non-linear pressure diffusion.
Earlier, Hummel and Müller (2009) analyzed synthetic clouds of 1D and 2D
microseismicity triggered by non-linear pore-pressure diffusion using an expo-
nential dependence of the diffusivity on the pore pressure (see also equations
(2.286)–(2.287)):
D(Pp ) = D0 exp(κ Pp ). (4.40)
Here D0 denotes diffusivity at vanishing pore pressure, and κ is the permeability

compliance. At the limit of vanishing κ the pressure relaxation will be described
by a linear diffusion equation. For the permeability compliance, Hummel and
Müller (2009) reported values from different literature sources ranging from 0 to
40 GPa−1 for sandstones and values up to 500 GPa−1 for fractured rocks. Millich
et al. (1998) state that for the KTB site the evolution of the in situ permeability with
effective pressure can be described by an exponential relation with a permeability
compliance value of 140 GPa−1 .
Hummel and Shapiro (2012) numerically considered power-law and exponen-
tial pressure dependencies of the diffusivity (4.21) and (4.40), respectively. They
restricted their computations to radially symmetric structures. They found that, for
a weak fluid–rock interaction (permeability enhancement at the injection source
is less than one order of magnitude), the fracturing domain diffusivity does not
differ significantly from the effective diffusivity empirically estimated by using
the seismicity behavior in r –t plots. If non-linearity is large, all three values,
the initial medium diffusivity, the fracturing domain diffusivity and the diffusiv-
ity obtained from the triggering front using the r –t-plot approach are significantly
different. However, in an order-of-magnitude approximation their results show that
the r –t-plot approach provides diffusivity estimates after the hydraulic stimulation
of the rock.
Hummel and Shapiro (2012) used boundary and initial conditions corresponding
to a point-like (or a small-radius injection cavity) pressure source of magnitude p0
applied instantaneously at t = 0 and remaining constant for all times t > 0. This
corresponds to a Dirichlet-type boundary condition (a constant injection pressure).
They used p0 = 10 Pa and correspondingly renormalized the realistic permeability
compliances κ to the range of values between 0 and 0.5 Pa−1 , so that the injec-
tion impact on the permeability would be equal to the one for an injection with
p0 = 10 MPa (a typical order of magnitude in the case of a reservoir stimulation).
This corresponds to values of the product κ p0 less than 5. For the index of non-
linearity n they used n = 0, 1, 2 and 3. In both exponential and power-law-diffusion
models they used the numerical value 1 for the quantity D0 . This corresponds to 1
m2 /s for the exponential model and 1 m2 /(sPan ) for the power-law diffusion.
Note that the initial and boundary condition problem we consider in this section
differs from previous problems. Here we define a constant-pressure boundary prob-
lem. In the previous section we considered boundary conditions corresponding to
pre-defined fluid-injection rates. To explore the behavior of the triggering front we
again start with the dimensional analysis.
For exponential diffusion model (4.40), the pore-pressure relaxation radius Rt
(we will again identify this radius with the radius of the triggering front) depends
on the quantities p0 , t, D0 , A0 and κ. Here A0 denotes the surface area of an

effective injection source cavity. The dimensions of significant physical variables
and parameters of the problem are:
L2 1
[ p0 ] = P, [t] = T, [D0 ] = , [A0 ] = L 2 , [κ] = , (4.41)
T P
where, L, T and P denote physical dimensions of length, time and pressure,
respectively. We can construct two dimensionless quantities, e1 and e2 :
D0 t
e1 = , e2 = κ p0 . (4.42)
A0
Then, the relaxation radius can be written in the following form:

Rt = D0 t f e (e1 , e2 ), (4.43)
where f e is a function of variables e1 and e2 .
In the case of power-law-diffusion model (4.21), the relaxation radius of the
pore-pressure perturbation depends on quantities p0 , t, D0 and A0 . The physical
dimensions of these quantities are
L2
[ p0 ] = P, [t] = T, [D0 ] = , [A0 ] = L 2 . (4.44)
T Pn
We can construct one dimensionless quantity only:
D0 t p0n
p1 = . (4.45)
A0
Hence, the relaxation radius has the form

Rt = D0 t p0n f p ( p1 ), (4.46)
where f p is a function of p1 .
For both the exponential and power-law diffusivity models, the strength of the
injection source is given by a surface-force density. This is the injection pressure
applied to the surface of the injection cavity. In one dimension, the injection source
represents a surface of a half-space. The pressure is distributed along an infinite
vertical plane intersecting the coordinate axes at the origin. Therefore, surface area
A0 of the effective injection source becomes infinitely large. As a result, the quan-
tity e1 goes to zero. Thus, for the exponential diffusion model, pressure fronts
(and, therefore, triggering fronts) are given by

Rt = D0 t f e (0, e2 ). (4.47)
For the power-law-diffusion model, triggering fronts in one dimension are given by

Rt = D0 t p0n f p (0). (4.48)
In equations (4.47) and (4.48) we assume that, for e1 → 0 and p1 → 0, the
function limits f e (0, e2 ) and f p (0) are finite. For both models describing non-
√
linear fluid–rock interaction relations (4.47) and (4.48) show a t-dependence of
the triggering front.
In two dimensions, the situation is similar. The pressure is distributed across the
surface of an infinitely extended cylinder so that A0 becomes again infinite. This
cylinder is normal to the plane in which pore-pressure diffusion takes place. As a
result, e1 → 0 and p1 → 0. Thus, triggering fronts for both the exponential-
diffusion model and the power-law-diffusion model are also described by relations
(4.47) and (4.48), respectively.
In contrast, in the case of pore-pressure diffusion in three dimensions, the injec-
tion pressure is distributed across the surface area of an effective injection-source
cavity of a finite radius a0 . In this case, the surface area is limited and does not
approach infinity. Moreover, we are interested in a solution for a point-like injection
source, that is, A0 → 0. Therefore, now we are looking for asymptotic behavior
of both functions f e and f p in the limits e1 → ∞ and p1 → ∞. For such
situations, it is more difficult to derive a relation for the triggering front. Accord-
ing to Barenblatt (1996), it is possible that the limits of f e and f p , if they exist,
will become either constant or power-law dependent. Hummel and Shapiro (2012)
numerically showed that, in the case of the exponential diffusivity model, the trig-
√
gering front still have a t-dependence. Therefore, the limit of f e is a constant.
However, for the power-law diffusivity model they observed a more general power
law of the temporal dependence of the triggering front. This indicates a power-law
dependence of f p (see also equation (4.46)):
bn −0.5
D0 t p0n
f p ( p1 ) ∼ , (4.49)
A0
where the exponent bn depends on the injection source and properties of the
medium.
Numerical results of Hummel and Shapiro (2012) (see Figures 4.20 and 4.21)
support these results of the dimensional analysis. They first computed the pres-
sure distribution for several numerical models. In the case of non-linear diffusion
(κ > 0 and n > 0) the hydraulic diffusivity strongly increases with pressure. As
a result, the pore-pressure profiles (Figure 4.20) penetrate deeper into the medium
compared to the linear diffusion. Owing to the geometrical spreading in models
with higher spatial dimension, the pore-pressure profiles are characterized by a
large pressure drop in the vicinity of the source. With increasing influence of the
non-linear fluid–rock interaction, the shape of the pore-pressure profiles tends to
change from a concave one to a more convex one. For the exponential diffusion
1D exponential diffusion, t = 100s 1D power−law diffusion, t = 100s
10 10
9 9
n=3
8 8
Pore pressure [Pa]

Pore pressure [Pa]
7 κ = 0 Pa−1 7
6 κ = 0.5 Pa−1 6 n=2
κ = 0.2 Pa−1
5 5
κ = 0.4 Pa−1
4 4 n=1
3 3
2 2 n=0
1 1
0 0
0 20 40 60 80 100 120 0 100 200 300 400 500 600
Radial 2D exponential diffusion, t = 100s Radial 2D power−law diffusion, t = 100s

10 10
9 9
8 8
Pore pressure [Pa]
Pore pressure [Pa]

7 7
6 6
5 5 n=3
−1
4 κ = 0.5 Pa 4
n=2
3 κ = 0 Pa−1 3
2 κ = 0.2 Pa−1 2 n=1
κ = 0.4 Pa−1
1 1 n=0
0 0
0 10 20 30 40 50 60 0 50 100 150 200
Radial 3D exponential diffusion, t = 100s Radial 3D power−law diffusion, t = 100s
10 10
9 9
8 8
Pore pressure [Pa]
Pore pressure [Pa]
7 7
6 6
5 5
4 4
−1
3 κ = 0.5 Pa 3
κ = 0.4 Pa−1
2 2
κ = 0.2 Pa−1 n=3
1 κ = 0 Pa−1 1 n=0
n=1 n=2
0 0
0 2 4 6 8 10 12 0 5 10 15 20 25 30
Figure 4.20 Snapshot profiles of the pore-pressure diffusion (at the moment 100 s
after the injection start with p0 = 10 Pa) in numerical models with exponential
(left-hand column) and power-law (right-hand column) pressure dependence of
the hydraulic diffusivity (equations (4.40) and (4.21), respectively). The profiles
are shown for various κ and n. From the top to bottom, respectively, the profiles
correspond to 1D, 2D and 3D media. (Modified from Hummel and Shapiro, 2012.)
model, the pressure profiles are characterized by a smooth transition at the pres-
sure heads (as in the linear-diffusion case). For the power-law model, the pressure
drops to zero abruptly.
Further, Hummel and Shapiro (2012) have numerically simulated microseismic
clouds in their numerical models of pore-pressure diffusion. Figure 4.21 shows
−1
1D exponential diffusion, κ = 0.5 Pa 1D power−law diffusion, n = 3
120 600
7345 events 42268 events
Deff = 6.4902 m2/s Deff = 212.5881 m2/s
100 500
Dh = 9.3891 m2/s Dh = 241.2881 m2/s
Distance [m]
Distance [m] 80 400
60 300
40 200
20 100
0 0
0 20 40 60 80 100 0 20 40 60 80 100
Time [s] Time [s]
−1
Radial 2D exponential diffusion, κ = 0.5 Pa Radial 2D power−law diffusion, n = 3
70 400
60 Deff = 2.5011 m2/s 350 Deff = 99.0172 m2/s
Dh = 3.3914 m2/s 300 Dh = 41.2089 m2/s
50
Distance [m]
Distance [m]
250
40
200
30
150
20
100
10 50
0 0
0 20 40 60 80 100 0 20 40 60 80 100
Time [s] Time [s]
−1
Radial 3D exponential diffusion, κ = 0.5 Pa Radial 3D power−law diffusion, n = 3
50 50
45 45 Deff = 1.9206 m2/s
Deff = 1.061 m2/s
40 Dh = 1.6861 m2/s 40 Dh = 0.77558 m2/s
35 35 r = (136.0611*t)0.36171
Distance [m]
Distance [m]
30 30
25 25
20 20
15 15
10 10
5 5
0 0
0 20 40 60 80 100 0 20 40 60 80 100
Time [s] Time [s]
Figure 4.21 The r –t plots of synthetic microseismicity together with the results
of the diffusivity analysis for the exponential (left-hand column) and power-law
(right-hand column) diffusion models in 1D (top panels), 2D (middle panels) √
and 3D (bottom panels) media. One can see that for 1D and 2D media, the t-
dependent triggering front (solid parabola) is the envelope of the microseismic
event clouds. This is also the case for the 3D diffusion in an exponential model.
In the 3D power-law model the hexagram-marked curve represents a function
r ∝ t bn describing the envelope of seismicity. (Modified from Hummel and
Shapiro, 2012.)
selected r –t plots from their publication. It is seen that for the exponential dif-
√
fusion model the envelope of the microseismic clouds is a t parabola. This
behavior of the triggering front is influenced neither by the non-linearity nor by
√
the spatial dimension. The t-dependence agrees with the results obtained from
our dimensional analysis. Figure 4.21 also shows several results of Hummel and
Shapiro (2012) for the power-law model. The results are in a good agreement with
our dimensional analysis too. Indeed, envelopes of 1D and 2D synthetic micro-
√
seismic clouds also have t-dependence. For the 3D case the situation changes.
√
A t-parabola (solid curve) does not represent an envelope of the microseis-
mic cloud. We observe a more general power-law time dependence t bn (see also
equations (4.46) and (4.49)). Hummel and Shapiro (2012) obtained the follow-
ing exponents: for n = 1 bn ≈ 0.40, for n = 2 bn ≈ 0.37 and for n = 3 bn
≈ 0.36. These results agree very well with equation (4.28) from the previous sec-
tion. Indeed, for a constant fluid-injection rate (i = 0) in three dimensions (d = 3)
this equation provides a power-law temporal behavior of the triggering front t bn
with very close exponents:
n+1
bn = . (4.50)
3n + 2
Hummel and Shapiro (2012) attempted further to use synthetic microseis-
mic clouds for estimating hydraulic diffusivity in corresponding models. They
computed heuristic diffusivity by using equation (3.6) and compared it with
approximate estimates of the effective diffusivity of the stimulated medium. To
compute the effective diffusivity they used the following approach.
The fluid stimulation leads to a hydraulic diffusivity, which increases with time
and is heterogeneously distributed in space. At the last moment of the injection
a monotonic (non-decreasing) stimulation creates a maximum enhanced hydraulic
diffusivity. An effective-diffusivity value represents such an enhanced diffusivity.
The replacement of a heterogeneously distributed diffusivity by a single effective-
diffusivity value is designed in such a way that the position of a triggering front in
the replacing homogeneous medium would approximately coincide with its actual
position (see also Section 3.6.3).
For a particular synthetic microseismic event cloud, one subdivides a distance
r f e from the injection point to the farthest event at a given time t f e into equidistant
intervals dr . For time t f e and all distance samples r , one computes the corre-
sponding pore-fluid pressure p(r ; t f e ). This pore pressure is used to compute the
stimulated diffusivity at the point r and time t f e . Then, corresponding to equation
(3.50), an estimate of the effective diffusivity can be obtained by the following
integration:
rfe −1
dr
De f f (r f e ; t f e ) = r f e . (4.51)
0 D(r ; t f e )
This estimate taken at the last moment of the injection is assumed to be the effective
diffusivity of the stimulated structure. Hummel and Shapiro (2012) compared the
heuristic diffusivity estimates obtained from the triggering front to the effective dif-
fusivity values computed by using equation (4.51). They observed that the heuristic
diffusivity estimates and effective diffusivity values show good correlation for
indices of non-linearity n < 2 in one dimension and n < 3 in two dimensions. For
larger values of n, the heuristic estimates are approximately half as much as the
effective values. However, the order of magnitude of the estimates is still the same
compared to the effective values. Even in the case of very strong non-linear fluid–
rock interaction triggering front (3.6) still provides reasonable diffusivity estimates.
This can be also seen from the parabolic curves corresponding to the effective and
heuristic diffusivity estimates shown in Figure 4.21.
4.3 The model of factorized anisotropy and non-linearity

Now we can try to use the power-law non-linear diffusion model for an interpre-
tation of the Barnett Shale microseismic data set we considered in Section 4.1.6.
Shale and most other rocks associated with hydrocarbon reservoirs can be strongly
elastically and hydraulically anisotropic. The anisotropy controls the distribution of
microseismicity. Below we follow Hummel and Shapiro (2013) and discuss their
approach to account for non-linear anisotropic pore-pressure diffusion.
We make several strong assumptions. We first assume that the principal direc-
tions of the pressure-dependent tensor of hydraulic diffusivity remain unchanged
during the stimulation. We consider the pressure-dependent diffusivity tensor Dik
in its principal coordinate system. Then, this tensor is represented by a diago-
nal matrix with the principal components D11 , D22 and D33 . Furthermore, we
assume that the pressure dependence of any of these components can be expressed
by the same function f D (Pp ). For example, for both models considered in the
previous section this function is given by an exponential law or a power law.
Thus, we assume that the tensor of diffusivity can be described by the following
decomposition of its pressure-dependent and anisotropic parts:
⎡ ⎤
D011 0 0
D(Pp ) = ⎣ 0 D022 0 ⎦ f D (Pp ). (4.52)
0 0 D033
The pressure-dependent part describes an impact of the hydraulic stimulation on
the permeability. In general, the function f D (Pp ) can express an arbitrary func-
tional dependence on pressure. Hummel and Shapiro (2013) assumed a power-law
diffusivity: f D (Pp ) = (n + 1)Ppn . The matrix in (4.52) controls the directivity of
the stimulated hydraulic transport. Then we substitute (4.52) into (3.1) and obtain
the following form of the non-linear diffusion equation:
∂ Pp ∂ ∂
= D011 f D (Pp ) Pp
∂t ∂ x1 ∂ x1
∂ ∂ ∂ ∂
+ D022 f D (Pp ) Pp + D033 f D (Pp ) Pp . (4.53)
∂ x2 ∂ x2 ∂ x3 ∂ x3
Now we can apply the transformation approach described in Section 3.5. We use
here a slightly different normalization:
Dn Dn Dn
X 1 = x1 , X 2 = x2 , X 3 = x3 , (4.54)
D011 D022 D033
where

Dn = 3
D011 D022 D033 (4.55)
is a reference diffusivity. Note that the choice of Dn is a matter of convenience. For
example, in Section 3.5 we used Dn = Dii /3.
In this coordinate system, equation (4.53) becomes

∂ Pp ∂ ∂ ∂ ∂ ∂ ∂
= Dn f D (Pp ) + f D (Pp ) + f D (Pp ) Pp .
∂t ∂ X1 ∂ X1 ∂ X2 ∂ X2 ∂ X3 ∂ X3
(4.56)
Thus, in this coordinate system for a spherically symmetric source (e.g. a point
source or a spherical cavity) we can assume radial symmetry of the solution and
write this equation in spherical coordinates (see also (4.16)):
∂r 2 Pp ∂ ∂
= Dn f D (Pp )r 2 Pp . (4.57)
∂t ∂r ∂r
This model accounts for permeability variations due to the reservoir stimulation.
However, it assumes that the functional type of the pressure dependence of the
permeability is identical for any direction at any location. Under these conditions
a cloud of induced seismicity in the normalized coordinate system X 1 , X 2 , X 3 will
stay spherical at all times that elapse after the start of the stimulation.
The model of a factorized anisotropy and non-linearity can be applied to real
data in the following way. We first should compensate for the directivity of the
hydraulic transport. This can be done by re-normalization of the cloud from its
originally nearly ellipsoidal shape to a spherical shape. Characteristic scales of the
microseismic cloud can be estimated at the injection final moment t0 . It is con-
venient to rotate the event cloud into its principal coordinate system and then to
estimate the lengths of its principal axes. For the particular Barnett Shale case
study they are L x = 1121 m, L y = 393 m and L z = 80 m. In accordance with
(4.54) and (4.55), the scaling of the coordinates of seismic events is:
Ln Ln Ln
X1 = x , X 2 = x2 , X 3 = x3 , (4.58)
Lx Ly Lz
where

Ln = 3
Lx L y Lz. (4.59)
Hummel and Shapiro (2013) applied scaling (4.58) to the cloud shape corre-
sponding to several times elapsed after the injection start. Figure 4.22 shows that
T/23 T/1.53 T
500 500 500
Y [m]
Y [m]
Y [m] 0 0 0
Lint
–500 –500 –500
–500 0 500 –500 0 500 –500 0 500
X [m] X [m] X [m]
500 500 500
Lmax
Z [m]
Z [m]
Z [m]
0 0 0
–500 –500 –500

–500 0 500
–500 0 500 –500 0 500
X [m] X [m] X [m]
500 500 500
Z [m]
Z [m]
Z [m]
0 0 0
Lmin
–500 –500 –500

–500 0 500 –500 0 500 –500 0 500
Y [m] Y [m] Y [m]
+ original seismicity cloud + transformed seismicity cloud
Figure 4.22 Temporal evolution of the original (black) and transformed (gray)
event cloud. Vertical and horizontal scales are equal. Note that, for all three snap-
shots (here T denotes the complete injection duration and x, y and z are the two
horizontal and one vertical dimensions of the microseismic cloud, respectively),
scaling (4.58) has not changed. (Modified from Hummel and Shapiro, 2013.)
such scaling results in the spherical shape of the transformed cloud at all elapsed
times. This implies equivalent functional pressure dependence of the stimulated
permeability in different directions and thus supports the model of the factorized
anisotropy and non-linearity, i.e. equation (4.52).
Now one can use the triggering front to analyze spatio-temporal features of the
transformed event cloud. For such an analysis, Hummel and Shapiro (2013) applied
the approach that we discussed in the previous section. They considered a power-
law fitting function r (t) = (an t)bn ∗ 1 m for the triggering front. They obtained the
following estimates of the parameters: an ≈ 180 s−1 and bn ≈ 0.36. The corre-
sponding parabola r (t) approximates satisfactory the triggering front in the scaled
coordinate system. The values of the parameter bn approximately correspond to
the non-linearity index n in the range of 5–7. Then they simulated Pp (r, t) in the
case of a pressure source. For this they assumed a constant injection pressure equal
to the time average of the one measured in the experiment (approx. 8.4 MPa; see
Figure 4.23). In the numerical modeling this pressure was applied to the surface of
an injection cavity with radius 0.5 m. Then they adjusted the parameter D0 so that
(a) Engineering data used for modeling
12
Pinj Average Pinj
Pressure [MPa]
10
6
0 1 2 3 4 5 6
Scaled flow rate [m/s]
0.05 Scaled flow rate Average scaled flow rate
0.045
0.04
0 1 2 3 4 5 6
Time [h]
(b) n=5
10
Averaged scaled flow rate
Averaged injection pressure
8
Pore−fluid pressure [MPa]
4
1h 2h 3 h 4 h 5.4 h
0
0 50 100 150 200 250
Distance [m]
Figure 4.23 (a) Injection pressures and filtration velocities (calculated from the
injection rate for the spherical source of the radius of 0.5 m) from the hydraulic
fracturing of the Barnett Shale case study. (b) Numerically simulated time-
dependent pressure profiles based on the averaged injection pressures and on the
flow rates. Estimated values D0 = 3.4 · 10−32 m2 /(s Pa5 ) (note that this is just a
normalizing constant depending on n; it should not be interpreted independently;
here n = 5) and S −1 = 8.4 GPa correspond to the pore-pressure perturbation
penetrating up to 250 m into the medium. (After Hummel and Shapiro, 2013.)
the diffusion front reaches the size of the seismicity cloud at the time t0 = 5.4 h
elapsed after the injection start (see Figure 4.23). The radius of the transformed
cloud is approximately 250 m.
Afterwards, the boundary condition was changed to a fluid-flow source. This is
the Neumann boundary condition. In other words, on the surface of the injection
Medium permeability
102 n=5
n=6
n=7
n5: κ = 0.008 darcy
n6: κ = 0.007 darcy
n7: κ = 0.005 darcy
darcy 100
1h
2h
3h
4 h 5.4 h
10–2
0 50 100 150 200 250

Distance [m]
Figure 4.24 Snapshots of the medium permeability computed for the pressure
profiles like those shown in Figure 4.23. The permeability of the stimulated rock
(excluding the singular source vicinity) is of the order of 10−3 –10−1 darcy. In
comparison, the permeability (dashed lines) based on the diffusivity estimates
using the triggering front are in the order of 5 − 8 · 10−3 darcy. The permeabil-
ity of the virgin rock is in the range of nano-microdarcy. (After Hummel and
Shapiro, 2013.)
cavity the quantity −DS∇ Pp was given. Corresponding to the Darcy law, in an
isotropic medium, this quantity is equal to the filtration velocity q. Giving the
averaged filtration velocity at the injection cavity surface approximately equal to
0.046 m/s (Figure 4.23) and adjusting now the value of the unknown storage coeffi-
cient S so that the pressure profile coincides with the one in the previous simulation,
one obtains S −1 = 8.4 GPa (this value corresponds to n = 5; see Figure 4.23).
Now we have all necessary parameters to reconstruct the spatio-temporal evo-
lution of the stimulated permeability. Examples of such reconstructions are shown
in Figure 4.24. Here we neglected possible pressure dependences of the storage
coefficient and of the density of the treatment fluid. Hummel and Shapiro (2013)
show some indications of these dependencies in the case study under consideration.
Nonetheless, this modeling example shows the possibility of a reasonable charac-
terization (which is in agreement with engineering and microseismic data) of the
stimulated reservoir permeability and of the storage coefficient using the non-linear
diffusion approach.
5
Seismicity rate and magnitudes
Magnitudes M of fluid-induced seismicity are usually in the range −3 < M < 2.

Nevertheless, especially for long-term injections with durations of months or years,
earthquakes with larger magnitudes (M = 4 or even larger) have been observed
(Ake et al., 2005; Majer et al., 2007). Fluid-induced earthquakes with M from 3
to 4 occurred at several Enhanced Geothermal Systems (EGS) like those of Basel,
Cooper Basin, The Geysers Field and Soultz (Giardini, 2009; Majer et al., 2007;
Häring et al., 2008; Dyer et al., 2008; Baisch et al., 2009). Smaller but still
perceptible events can be also observed by hydraulic fracturing of hydrocarbon
reservoirs.
Induced seismic hazard is a topic of significance in the shale-gas indus-
try (National Research Council, 2013). Its understanding is of a considerable
importance for mining of deep geothermic energy. It is of significance for CO2
underground storage (see Zoback and Gorelick, 2012) and possibly also for other
types of geo-technological activities (see Avouac, 2012).
Identifying parameters that control magnitudes and their statistics is a key
point for evaluating the seismic hazard of fluid injections. Similarly to the
tectonic seismicity, statistics of the induced seismicity can be rather well
described by the Gutenberg–Richter frequency–magnitude distribution (Shapiro
et al., 2007, 2010, 2011; Shapiro and Dinske, 2009a,b; Dinske and Shapiro, 2013).
However, large-magnitude events deviate from it (Shapiro et al., 2011). In this
chapter we analyze the influence of fluid injections on the frequency–magnitude
statistics of induced events. We start with a model of point-like independent
seismic events. This model describes the statistics of numerous small-magnitude
earthquakes well. The model allows us to formulate a simple description of the
seismicity rate and to introduce parameters quantifying the seismo-tectonic state
of a fluid-injection site. One such useful parameter is the seismogenic index. This
helps to predict the probability of given-magnitude events. However, the model of
point-like events tends to overestimate the probability of significant magnitudes. In
201
202 Seismicity rate and magnitudes
the second half of the chapter we consider statistics of large-magnitude events. We

take the finiteness of rupture surfaces of such earthquakes into account and con-
sider the influence of the finiteness of stimulated-rock volumes on this statistics.
We also address the issue of the maximum-magnitude induced earthquake.
5.1 The model of point-like induced events

Shapiro et al. (2007) proposed a model that allows us to calculate the expected
number of events with a magnitude larger than a given magnitude value M. This
model will enable us to identify the main factors that affect the magnitude prob-
abilities. Later we will see that this model is rather successful for describing
the statistics of numerous small-magnitude events. However, it overestimates the
number of large-magnitude earthquakes.
The model assumes a non-linear pore-pressure relaxation as a triggering mech-
anism of induced seismicity. We continue to consider a point-like pressure source
in an infinite, homogeneous, permeable, porous continuum and repeat our argu-
mentation for equation (3.74). As a result of a fluid injection and the consequent
process of pressure relaxation, the pore pressure Pp will change throughout the
pore space. We assume that a random set of pre-existing cracks (defects) is statisti-
cally homogeneously distributed in the medium and is characterized by the volume
concentration N . For simplicity we assume that the cracks do not mutually inter-
act. Each of these cracks is characterized by an individual critical value C of the
pore pressure necessary in accordance with the Mohr–Coulomb failure criterion
for the occurrence of an earthquake along such a defect. This critical pressure C is
randomly distributed on a set of pre-existing cracks. If C is high we speak about a
stable pre-existing fracture. If C is low we mean a fracture close to its failure. Sta-
tistical properties of C are assumed to be independent of spatial locations (i.e. C(r)
is a statistically homogeneous random field). If at a given point r of the medium
(with a pre-existing crack there) pore pressure Pp (t, r) increases with time, and at
time t0 it becomes equal to C(r), then this point will be considered as a hypocenter
of an earthquake occurring at this location at time t0 . For simplicity, we assume
that no earthquake will be possible at this point again. This is equivalent to an
assumption that phenomena leading to recharging critical cracks, such as thermal
relaxation, stress corrosion, viscoelastic tectonic deformation and other effects that
may be also related to the rate- and state-dependent friction, described by Dieterich
(1994), Segall and Rice (1995), Segall et al. (2006) and mentioned in our Section
1.2.2, are much slower than the diffusional process of the pore-pressure relaxation.
Under these assumptions the probability of an earthquake occurring in the time
interval from the injection start till a given time t and at a defect located at a given
point r will be equal to Wev (C(r) ≤ Maximum{ p(t, r)}). This is the probability of
the critical pressure being smaller than the maximum pore pressure at r in the time
interval from the injection start till the time t. If the pore-pressure perturbation
caused by the fluid injection is a non-decreasing function, which is the case for
non-decreasing borehole injection pressures, then this probability will be given by
equation (3.74). Rothert and Shapiro (2007) have shown that C is usually of the
order of 103 –106 Pa (see also Section 3.8), and f (C) can be roughly approximated
by a boxcar function f (C) = 1/(Cmax − Cmin ) if Cmin ≤ C ≤ Cmax and f (C) = 0
elsewhere (see Section 3.8). Usually Cmax is of the order of several MPa. It is orders
of magnitude larger than Cmin . Thus, f (C) ≈ 1/Cmax . We also assume that Cmax
is larger than the pressure perturbation (excluding maybe a small vicinity of the
source). Then,
Wev = p(t, r)/Cmax . (5.1)
Therefore, the event probability is proportional to the pore-pressure perturbation.
In turn, a spatial density of events is proportional to the event probability. Thus, the
spatial event density must be also proportional to the pore-pressure perturbation.
Such a distribution of microearthquake spatial density is indeed observed in reality
(see Section 3.9).
5.1.1 Event number and event rate during a monotonic injection

It is apparent that the probability of an event with a magnitude larger than a given
value is an increasing function of the total event number. Thus, first we compute
the complete number of events induced from the injection start till a current time
moment t. The product of the crack bulk concentration N and a spatial integral of
probability (5.1) will give us the necessary quantity:
∞
Ad N
Nev (t) = p(t, r )r d−1 dr. (5.2)
Cmax a0
Here the spatial integration is similar to the one of equations (4.18)–(4.19). From
equation (4.20) we further obtain:
N m c (t)
Nev (t) = , (5.3)
Cmax ρ0 S
where m c (t) is a fluid mass injected until time t and S is the storage coefficient.
This result is a consequence of the conservation of the fluid mass. Thus, it is valid in
the very general case of non-linear fluid–rock interaction. We see that the number
of events induced is proportional to the injected fluid mass. Note that an important
validity condition for this statement is the non-decreasing pore-pressure perturba-
tion in the whole space. If we neglect possible changes of the fluid density then this
result will simplify correspondingly:
t
N Ad N
Nev (t) = Q i (t)dt = Q 0 t i+1 , (5.4)
Cmax S 0 Cmax
where in the last expression a power-law injection rate (parameterized after
equation (4.20)) was assumed.
Finally, in the case of a constant injection rate Q i (t) = Q I = S Ad Q 0 we obtain
QI Nt
Nev (t) = . (5.5)
Cmax S
This result shows that the cumulative event number is growing proportionally to
the injection time. In other words, if the injection rate is constant, the events rate

Nev (t) will also be a constant equal to
∂ Nev QI N
Nev (t) ≡ = . (5.6)
∂t Cmax S
The quantity
Cmax
Ft = (5.7)
N
depends on seismotectonic properties of the injection region only. Following
Shapiro et al. (2007) we will address this quantity as a “tectonic potential.” In
our simple model, the tectonic potential is defined by a bulk concentration of
pre-existing cracks (i.e. defects or, in other words, potential rupture surfaces for
induced events) and the upper limit of their critical pressure Cmax . Note that 1/Cmax
defines an average probability density of critical pressures. Note also that the tec-
tonic potential has physical units of energy. Such a relatively simple expression
of the tectonic potential is due to simplicity of our model. In reality, Ft can be a
more complex function of stress state, tectonic history, rheology, lithology, hit flow,
seismicity and possibly other parameters of the injection site.
According to (2.200), (2.220) and (2.221), by a constant injection rate and a
linear pressure diffusion, we have Q I /S = 4π Dp0 a0 , where p0 is the injec-
tion pressure. This relation provides us with an alternative interpretation of
equation (5.5):
Nev (t) = 4π Dp0 a0 t/Ft . (5.8)
Equations (5.5) and (5.8) show that the probability of seismicity increases with the
volume of the injected fluid, with the reciprocal tectonic potential, with the stiff-
ness 1/S of fluid saturated rocks (for example, with decreasing porosity, increasing
fluid stiffness, increasing rock consolidation, etc.) and with the diffusivity of the
medium.
To compute the number of events with magnitudes larger than a given one, we
must introduce some assumptions about the magnitude statistics. We assume that
the probability W≥M of an event having a magnitudes larger than M is independent

of the number of events. This means that the ratio of a number of events with
magnitudes larger than a given one to the complete number of events Nev is, in
average, a constant independent of the value of Nev . Then, we obtain from (5.3):
m c (t)
N≥M (t) = W≥M . (5.9)
Ft ρ0 S
This rather general result has the following interesting consequence. In the log-
arithmic scale we obtain (note that here we are using the International Standard
notation lg ≡ log10 and ln ≡ loge )
Ad Q 0
lg N≥M = lg W≥M + lg m c − lg Ft ρ0 S = lg W≥M + lg + (i + 1) lg t, (5.10)
Ft
where the last equality is obtained for a power-law injection rate considered in
the previous chapter. This equation means that, in a logarithmic plot, curves of
functions N≥M (t) for different M must be mutually parallel. This is a consequence
of the assumption of independence of magnitude probability of event numbers.
Figures in the following section will confirm this statement. Moreover, in the case
of power-law injections, on bi-logarithmic plots N≥M versus t, theoretical curves
N≥M (t) are straight lines with proportionality coefficients i + 1.
To be more specific, we assume further that the induced seismicity obeys
Gutenberg–Richter statistics (Gutenberg and Richter, 1954). Broad seismological
experience shows that these statistics describe global as well as local regional tec-
tonic seismicity (see, for example, Shearer, 2009; and Turcotte et al., 2007). The
Gutenberg–Richter statistics of an induced seismicity might be a consequence of a
power-law-type size distribution of pre-existing defects (Langenbruch and Shapiro,
2014; see also our Section 5.2.4). The Gutenberg–Richter magnitude scaling means
that the probability W≥M of events with magnitudes larger than M is given as
follows:
lg W≥M = a − bM, (5.11)
where a and b are the so-called Gutenberg–Richter a- and b-parameters (or val-
ues). They are regional seismicity constants. Note also that a = lg W≥0 . Thus,
from (5.9), we obtain
m c (t) Ad
N≥M (t) = 10a−bM = 10a−bM Q 0 t i+1 . (5.12)
Ft ρ0 S Ft
This result clarifies which characteristics of rocks and of injections define mag-
nitudes of induced earthquakes. The probability of significant events increases
with an injection duration, with the strength of an injection source and with the
reciprocal tectonic potential of an injection site (i.e. with the concentration of
critical cracks multiplied by an average of critical-pressure probability density).
The number of events with magnitude larger than a given one is proportional to the
cumulative mass of the injected fluid. This equation also shows a temporal scaling
of this number, t i+1 . It is controlled by the scaling i of the injection rate. Note also

a simple rule for the event rate N≥M (t) ≡ ∂ N≥M (t)/∂t:
N m i (t) Q i (t) Ad (i + 1)
N≥M (t) = 10a−bM = 10a−bM = 10a−bM Q 0 t i . (5.13)
Cmax ρ0 S S Ft Ft
Therefore, the event rate is proportional to the mass rate of the injection or, approxi-
mately, to the volumetric injection rate. A temporal scaling of N≥M is convenient to
consider in a bi-logarithmic scale. For the logarithm of the event number, equation
(5.12) results in the following:
m c (t) Ad Q 0
lg N≥M (t) = lg − bM + a = lg + (i + 1) lg t − bM + a. (5.14)
Sρ0 Ft Ft
Again, in a bi-logarithmic coordinate system in the case of an injection rate scaled
as t i , N≥M as a function of time tends to be a straight line with a proportionality
coefficient i + 1.
5.1.2 Seismicity rate after termination of an injection

In the following sections we will compare our rather general conclusions on the
seismicity rate of non-decreasing injections with real data. In this section we first
briefly comment on the seismicity decay after a terminated injection.
Predicting the seismicity rate after the termination of an injection requires sig-
nificantly more complete knowledge (or more assumptions) about the seismicity
triggering process. Assuming that linear pore-pressure diffusion is the governing
mechanism of seismicity triggering and assuming a constant injection rate into
a homogeneous porous medium, Langenbruch and Shapiro (2010) formulated an
analytical solution and calculated numerically the seismicity rate after an injection
stop. They found that the decay rate of the induced seismicity is similar to the
Omori law, which describes the decay rate of aftershock activity after tectonically
driven earthquakes. For natural earthquakes it is usually claimed that the Omori
exponent is close to 1. However, values in the range 0.3–2 have also been reported.
Two models frequently used to explain such behavior of aftershocks are the rate-
and state-dependent behavior of the friction coefficient and the stress corrosion of
sub-critical cracks (Kanamori and Brodsky, 2004).
Langenbruch and Shapiro (2010) found this exponent to be larger than 1 for the
case of the fluid-induced seismicity. Let t0 denote the duration of a fluid injection
and let τ0 = t/t0 denote a normalized time after the termination of the injection
(τ0 ≥ 1). Langenbruch and Shapiro (2010) propose the following approximation
of the seismicity rate after the termination of the injection (at least for τ0 = O(1)):
−p
Nev (t) ≈ Nev (t0 )τ0 d , (5.15)
where the Omori exponent pd characterizes the decay rate, and (see also (2.220)
and (2.221))
√
Cmin Dt0
pd − 2 ∝ . (5.16)
Cf
In other words, if the minimum critical pressure is very low and τ0 = O(1) then
the Omori exponent for the seismicity induced under conditions described above
will approach 2. Further, their analytic results show that, for τ0 1, the Omori
exponent approaches 1. However, their estimations for the case studies Fenton Hill
and Soultz (in the range of τ0 of the order of 1) provided as high values of the
Omori exponent as 7.5 and 9.5, respectively. This corresponds to Cmin of the order
of 5000 Pa. Thus, the decay rate of induced seismicity depends on the fracture
strength. Moreover, the values show that the presence of unstable fractures (with
the vanishing criticality C) results in an increase of the seismicity rate shortly after
an injection stop (compare Figures 5.1 and 5.2). This also leads to a probability
increase for large-magnitude events shortly after the injection termination. Later in
this chapter we will observe that this effect can be additionally enhanced by the
growing scale of the finite stimulated volume with the time elapsed after the start
of the injection. In the case of Cmin = 0, one obtains
4π DC f
Nev (t0 ) ≈ ν I ≡ . (5.17)
Ft
We call this quantity the reference seismicity rate.
injection phase post−injection phase

1.2
Reference rate R0
1
Cmin = 1000 Pa
0.8 Cmin = 10 000 Pa
Omori law (p = 1.8)
Rs(t)
0.6
Omori law (p = 3)
0.4 Omori law (p = 10)
0.2
0
0 1000 2000 3000 4000 5000 6000 7000 8000 9000
t [s]
Figure 5.1 Seismicity rate in the case of stable pre-existing fractures (Cmin is
finite and significant). The rate is normalized to the reference seismicity rate ν I
defined in (5.17). Parameters of the model are D = 1 m2 /s, t0 = 2000 s, p0 =
1 MPa, a0 = 4 m, Cmax = 1 MPa. The solid lines correspond to Cmin = 0 Pa,
Cmin = 1000 Pa and Cmin = 10 000 Pa (from the upper to the lower curve). The
dashed lines show the modified Omori law with pd = 1.8, 3.0 and 10.0. (After
Langenbruch and Shapiro, 2010.) A black and white version of this figure will
appear in some formats. For the color version, please refer to the plate section.
1.2
Reference rate R0
1
Cmax = 105 Pa
0.8
Cmax = 104.5 Pa
Rs(t)
0.6 Cmax = 104 Pa
0.4
0.2
0
0 1000 2000 3000 4000 5000 6000 7000 8000 9000
t[s]
Figure 5.2 Seismicity rate in the case of unstable pre-existing fractures (Cmin =
0). The rate is normalized to its maximum value, given by the reference seismicity
rate ν I . Parameters of the model are D = 1 m2 /s, t0 = 2000 s, p0 = 1 MPa,
a0 = 4 m. The solid lines correspond to Cmax = 106 Pa, Cmax = 105 Pa,
Cmax = 104.5 Pa and Cmax = 104 Pa (from the upper to the lower curve). The
arrows denote the time of maximum probability to induce an event with significant
magnitude. (After Langenbruch and Shapiro, 2010.) A black and white version of
this figure will appear in some formats. For the color version, please refer to the
plate section.
Note that all these observations can change significantly in the case of a non-
linear fluid–rock interaction. In contrast, features of the seismicity rate during non-
decreasing injection time intervals are governed mainly by the mass conservation
law and thus are more universal. In the next section we return back to seismicity
rate of active injection phases.
5.1.3 Case studies of magnitude distributions

Here we discuss the behavior of N≥M as a function of injection time for some
borehole injections: at Ogachi (Japan), in Paradox Valley (USA), in Cooper Basin
(Australia), at Basel (Switzerland), in Cotton Valley (USA) and in Barnett Shale
(USA). Note that for our discussion we use only those time intervals of these exper-
iments which approximately correspond to power-law temporal dependencies of
their injection rates, t i .
We restrict our analysis to magnitude ranges that are possibly weakly influenced
by features of observation systems, of registration and of processing. The so-called
completeness magnitude Mc is an important characteristic of seismicity catalogs.
It is defined in the following way. It is assumed that all events with magnitudes
M ≥ Mc are recorded by the observation system and registered in data sets.
First, we compare the number of events as a function of time as predicted
by equation (5.14) with observations from data sets at injection sites Ogachi (a
geothermal system) and in Paradox Valley (a salt-water disposal). Both injections
correspond approximately to the assumptions made above, and induced a suffi-
ciently large number of earthquakes (larger than 100). During an experiment at
Ogachi 1991
104
flow rate
pressure
103 M > –2.5
M > –2.0
NE VM>M0
M > –1.5
102
101
1 5 10 30
Time (days)
Figure 5.3 N≥M as functions of injection time for the Ogachi 1991 experiment.
The points are observed cumulative numbers of earthquakes with magnitudes
larger than the indicated ones. The straight line has the proportionality coeffi-
cient 1, predicted by equation (5.14). The curves show the injection pressure (the
lower line in the time range 1–10 days) and the injection rate (the upper line in the
time range 1–10 days). (Modified from Shapiro et al., 2007.) A black and white
version of this figure will appear in some formats. For the color version, please
refer to the plate section.
the Ogachi geothermal site in 1991, a volume of more than 10 000 m3 of water
was injected at a depth of 1000 m into hard rocks (granodiorite). The pressure
remained relatively stable throughout the experiment (see Figure 5.3). A micro-
seismic event cloud of about 500 m thickness and 1000 m length with nearly 1000
detected events was stimulated (Kaieda et al., 1993). The magnitudes were deter-
mined by measuring velocity amplitudes and, alternatively, seismogram oscillation
durations (Kaieda and Sasaki, 1998).
Magnitude statistics has been biased by the performance of the observation sys-
tem and by processing in the magnitude ranges M <−2.5 and M >−1.5. Here
Mc = −2.5 is an estimate of the completeness magnitude. However, there were
systematic errors in magnitude measurements above M = −1.5. When the injec-
tion pressure is nearly constant, the functions N≥M (t) are nearly linear in the
bi-logarithmic plot. The distances between lines corresponding to different mag-
nitudes M are regularly distributed and time independent. These two features are
as predicted by equation (5.14). The steps between the lines can be used to estimate
the b value of the Gutenberg–Richter law.
At Paradox Valley, the injection was carried out in several phases between 1991
and 2004 in order to reduce salinity in the Colorado River (Ake et al., 2005). The
brine was injected into a fractured limestone formation at a depth of more than
4 km. The injection became regular in 1996, but still showed some fluctuations
and a 20-day shut down every six months since the year 2000. The microseismic
event cloud extended to more than 15 km from the injection borehole. About 4000
events with a magnitude larger than −0.5 were induced. The largest event had a
magnitude 4.3. The completeness magnitude Mc was close to 0.5. Though being
one order of magnitude larger on the spatial scale and two orders of magnitude
larger on the temporal scale, the microseismic activity in Paradox Valley shows the
same features as described above for the Ogachi case study (see Figure 5.4).
The magnitude distributions observed in Figures 5.3 and 5.4 agree rather well
with (5.14), which predicts a linear relation between lg N (M, t) and lg(t). Sig-
nificant deviations seen in Figure 5.4 are explained by a strong irregularity of the
injection pressure at the Paradox Valley site.
Our model and its main consequence, equation (5.14), also provide a conve-
nient frame for comparison seismicity induced in different experiments. Figure 5.5
shows a common plot of magnitude frequencies as functions of time for the both
regions and three injection experiments. We can clearly see that the most important
Paradox Valley –1996 –2004

105
pressure
104
M > 0.5
103
NEVM >M0
M > 1.0
M > 1.5
102
M > 2.0
101
100
5000 1000 2000 3000
Time [days]
Figure 5.4 Pressure (irregular black line) and distribution of earthquake magni-
tudes for the Paradox Valley brine injection experiment. The straight line shows
the theoretical proportionality coefficient 1 given by (5.14) for a constant-rate
continuous-in-time fluid injection. (Modified from Shapiro et al., 2007.)
104
Ogachi 1991 –2.5
Ogachi 1993
Paradox Valley 1996–2004 –2.0
–1.5
103
NEVM >M0 –1.0
–0.5 0.5
2 1.0
10
1.5
2.0
101 –1
100 101 102 103 4
10 10
Time (days)
Figure 5.5 A combined plot of numbers of events with magnitudes larger then
given ones as functions of injection durations at Ogachi and at Paradox Valley.
Thin dashed lines correspond to equation (5.14) with i = 0. (Modified from
Shapiro et al., 2007.) A black and white version of this figure will appear in some
formats. For the color version, please refer to the plate section.
factor influencing the appearance of the seismicity on this plot is the sensitivity of
observational systems. For example, the surface-based registration system of Para-
dox Valley is not able to detect small-magnitude events. However, the influence of
this factor can be taken into account by extrapolation and construction of additional
lines corresponding to magnitudes of interest. The predictive potential of such a
plot can be seen. For example, it is clear that injection durations at the Ogachi site
are at least two orders of magnitude shorter than injection durations necessary to
induce events of M = 2 and larger. It is also seen that the experiment of Ogachi
1991 has approximately the same tectonic potential as the Paradox Valley injec-
tion. Figure 5.5 also shows the data from another injection experiment at Ogachi
performed in 1993. The magnitude distributions exhibit a similar behavior as for
the experiment in 1991. However, one can see that the later experiment is char-
acterized by a slightly larger tectonic potential. This is possibly due to criticality
releasing by the first injection experiment, which led to an effective decrease of the
concentration of critical cracks (one can also consider this as an increase in Cmax ).
Thus, the model proposed here can help to optimize a design of injection exper-
iments. For example, it can help to decide which parameters should be changed
(e.g. the injection rate or duration) in order to reduce the seismogenic risk of the
stimulation (see also the next section).
Further, we compare the data related to EGS projects in Australia at the Cooper
Basin and in Switzerland at Basel. Cooper Basin is located in South Australia.
We consider here the microseismicity induced in the borehole Habanero-1 (Soma
et al., 2004), which was drilled down to a depth of 4421 m. In its last 754 m
it penetrated granite. The open-hole section was below the depth of 4115 m.
However, the main outflow zone was at a natural fracture at the depth of 4254 m.
Basel is located in the Upper Rhine Graben (Dyer et al., 2008). The borehole
Basel 1 was drilled till the depth of approximately 4800 m and penetrated more than
2 km of granite in its lower part. The open-hole section corresponds approximately
to the last 400 m of the borehole.
Figures 4.16 and 4.17 show r –t plots of microseismic events induced in the
Cooper Basin and Basel experiments. Solid parabolic lines provide linear diffusion
approximations of microseismic cloud expansions (4.3) with hydraulic diffusivities
of the order of 0.4 m2 /s and 0.06 m2 /s, respectively. These approximations are
quite rough. However, in both case studies the estimates of hydraulic diffusivities
provided by linear diffusion approximations are still very realistic and correspond
well to hydraulic properties of naturally fractured granite.
We see (Figure 5.6) that the Cooper Basin injection period can be characterized
by a roughly constant flow rate. In contrast, the injection at Basel is characterized
by an increasing fluid flow. In the first case we can approximate the injection rate
exponent i by 0. In the second case, we can roughly approximate i by 1. This
corresponds to a smoothing of the real stepwise-increasing flow rate. Therefore,
corresponding to equation (5.14), we expect an approximately linear dependence
between lg N≥M and lg t in both cases. However, the proportionality coefficient of
this dependence is equal to i + 1. Thus, in the case of the Cooper Basin injection
this linear dependence should be characterized by the proportionality coefficient
1. In contrast, the Basel injection should be characterized by the proportionality
80 80 80 80
pressure flowrate pressure
Pressure [MPa]
Pressure [MPa]
Flow Rate [1/s]
Flow Rate [1/s]

60 60 60 flowrate 60
40 40 40 40
20 20 20 20
0 0 0 0
0 1 2 3 4 5 6 7 8 9 10 0 1 2 3 4 5 6
Time [days] Time [days]
3 3
Magnitude Mw
2
2
Magnitude
1
0 1
–1
0
–2
–3 –1
0 1 2 3 4 5 6 7 8 9 10 0 1 2 3 4 5 6
Figure 5.6 Water injection rate and pressure as functions of injection time as well
as observed microearthquake magnitudes in crystalline rocks of Cooper Basin
(left; the data are courtesy of H. Kaieda, CRIEPI) and of Basel (right; the data are
courtesy of U. Schanz and M. O. Häring, Geothermal Explorers). (After Shapiro
and Dinske, 2009b.)
104 104
103
103
NEV (Mw > M0)

NEV (M > M0)
102
M0 = 0.5
M0 = –0.5
102
M0 = 0.0 M0 = 1.0
101
M0 = 0.5
M0 = 1.0
0
101 1 5 10
10
1 5 10
Figure 5.7 Left: numbers of events with magnitudes larger than given ones as
functions of injection durations in the Cooper Basin. The thin dotted line has
proportionality coefficient 1. Right: numbers of events with magnitudes larger
than given ones as functions of injection durations in Basel. The thin dotted line
has proportionality coefficient 2. Note that only slopes of the dotted lines (rather
then their exact locations) are essential here. (After Shapiro and Dinske, 2009b.)
coefficient 2 of the lg N≥M versus lg t dependency. Figure 5.7 shows that this is
approximately the case.
Two other case studies are hydraulic fracturing injections into tight gas reser-
voirs of Cotton Valley sandstones and of Barnett Shale. Their r –t diagrams (see
equations (4.5) and (4.13), respectively) correspond to the growth of a single,
millimeters-thick dominant hydraulic fracture in the case of the Cotton Valley
reservoir stimulation (Figure 4.18), and to the growth of a 3D fracturing domain
(with additional compliant porosity of the order of 0.01%) in the case of the Bar-
nett Shale example (Figure 4.15). Figure 5.8 shows magnitude distributions in time
for both case studies. In both cases the injection rate is approximately constant.
Therefore, a linear dependence between lg N≥M and the lg t in both cases should
be characterized by the proportionality coefficient 1. This is in agreement with
Figure 5.9.
The fluid–rock interaction at the geothermal boreholes we considered above
was, to first approximation, linear (i.e. only small injection-induced alterations of
hydraulic diffusivity on the reservoir scale). The fluid–rock interaction in the case
of hydraulic fracturing of tight gas reservoirs was definitely strongly non-linear
(i.e. due to hydraulic fracturing the hydraulic diffusivity was strongly – by several
orders of magnitude – increased; this corresponds to a large index of non-linearity
n and, consequently, is strongly non-linear equation (4.22)).
In spite of this principal difference between examples describing stimulations
of the geothermal and tight gas reservoirs we observe a very similar character
200 50 200 50
flow rate bh pressure injection rate pressure
Pressure [MPa]
Flow rate [1/s]
Flow rate [1/s]
150 45 150 45
Pressure [MPa]
100 40 100 40
50 35 50 35
0 30 0 30
0 0.5 1 1.5 2 2.5 3 0 1 2 3 4 5 6
Time [h] Time [h]
0 –1
–0.5 –1.5
Magnitude
–1 –2
Magnitude
–1.5 –2.5
–2 –3
–2.5 –3.5
–3 –4
0 0.5 1 1.5 2 2.5 3 0 1 2 3 4 5 6
Time [h] Time [h]
Figure 5.8 Water injection rate and pressure as functions of injection time and
observed microearthquake magnitudes in Cotton Valley tight sands (left; the
data are courtesy of J. Rutledge, LANL) and Barnett Shale (right; the data are
courtesy of S. Maxwell, Pinnacle Technology). (Modified from Shapiro and
Dinske, 2009b.)
103 103
102 102 M > –3.0

NEV (M >M0)
NEV (M >M0)
M > –2.0
M > –1.75
M > –2.75
101 101
M > –1.5
M > –2.5
100 100
1 5 10 1 5 10
Time [hours] Time [hours]
Figure 5.9 Left: numbers of events with magnitudes larger than given ones as
functions of injection durations in the Cotton Valley tight sands. The thin dotted
line has proportionality coefficient 1. Right: numbers of events with magnitudes
larger than given ones as functions of injection durations in Barnett Shale. The
thin dotted lines have the proportionality coefficient 1, correspondingly to equa-
tion (5.14) with i = 0. Note that only slopes of the dotted lines (rather then their
exact locations) are essential here. (Modified from Shapiro and Dinske, 2009b.)
of magnitude distributions in time. This is in good agreement with the theoreti-

cal model developed above. Indeed, our equations (5.3) and (5.9) show that the
cumulative event number Nev (t) induced till time t, as well as the number of events
with the magnitude larger than a given one, N≥M , are independent of the index
of non-linearity, n. Therefore, the magnitude distribution in time is independent

of the non-linearity grade of the fluid–rock interaction. These event numbers as
functions of time are only controlled by the cumulative mass of the injected fluid.
The larger this cumulative mass, the larger is the number of events induced at a
given location. The larger the number of such events, the higher is the probability of
large-magnitude events. Equally larger will be the total seismic moment of events
induced. This is also in good agreement with the observation of McGarr (1976)
that the total seismic moment of induced earthquakes increases with the volume
perturbations in rocks. In addition to this observation, our model shows that the
magnitude distribution in time is a function of the spatio-temporal redistribution of
the pore pressure in rocks caused by a fluid-mass injection. Such a pore-pressure
redistribution also continues after the termination of the injection. Thus, the model
above is able to predict magnitude distributions after the injection stops. However,
such a prediction requires more-specific knowledge about the pressure-relaxation
laws. The mass-conservation principle alone is not sufficient for such a task. In
addition, all four considered case studies support our hypothesis on the indepen-
dence of the magnitude probability density of event numbers. It is seen from the
fact that, for different magnitudes M, the lines of lg N≥M as functions of time are
nearly mutually parallel.
A significant difference in the geothermal and tight-gas case studies is observed
in the b-values characterizing the microseismicity. The b-values are equal to the
differences between the lines of lg N≥M divided by the differences of correspond-
ing magnitudes. The approximate values for the Cooper Basin, Basel, Cotton
Valley and Barnett Shale examples are 0.75, 1.5, 2.5 and 2.5, respectively. We
observe approximately the following tendency: the larger the non-linearity, the
larger the b-value. Therefore, hydraulic fracturing is characterized by a much
stronger dominance of small earthquakes in the common number of induced seis-
micity than geothermal stimulations. Later in this chapter we will see that this is
likely to be the geometric effect of finiteness of the stimulated rock volume.
5.1.4 Seismogenic index

In the previous section we found that the rate of induced seismicity is controlled
by the rate of the injected fluid mass. Our observations supported the theoretical
prediction that by monotonic-rate (non-decreasing-rate) injections the number of
events with magnitudes larger than a given one increases with time as the injected
fluid mass does. This rule seems to work independently of the non-linearity grade
of the triggering process. Here we show that, by using the seismicity rate of
such events and the fluid-injection rate, a parameter (seismogenic index) can be
derived that quantifies the seismotectonic state of the injection site. This index is
independent of injection parameters and depends only on tectonic features of the

injection location. It can be used to quantitatively compare tectonic situations at
injection sites in terms of a potential risk to induce an event of a significant magni-
tude. Along with injection parameters the seismogenic index permits us to estimate
the occurrence probability of a given number of such events during a given time
period. We will also estimate this index for several injection experiments.
We rewrite equation (5.14) in the following form (we neglect possible variations
of the fluid density and take the first part of the right-hand side of the equation):
lg N≥M − lg Q c (t) + bM = a − lg (Ft S), (5.18)
where Q c is the cumulative injected volume and M is an arbitrary event magnitude.
The quantity on the left-hand side of this equation is experimentally measurable.
It depends on the injection parameters and on the induced seismicity. We are
concentrating now on the quantity defined by the right-hand side of this equation:
N
≡ a − lg(Ft S) = a + lg . (5.19)
Cmax S
This quantity is independent of the injection time. Neither does it depend on any
other injection characteristics. It is completely defined by the seismotectonic fea-
tures of a given location. We will call it the seismogenic index. The larger this
index, the larger a probability of significant magnitudes. If we know the - and
b-values, we will be able to predict the temporal distribution of magnitudes of
injection-induced events. In the case of a monotonic injection, for an arbitrary
injected cumulative volume Q c (t) the expected number of events with magnitudes
larger than M is
lg N≥M = + lg Q c (t) − bM. (5.20)
It is difficult to theoretically calculate because of some unknown parameters
(e.g. N ). However, can be estimated by using equation (5.18) and parame-
ters of the seismicity induced by an injection experiment at a given location (see
also Shapiro et al., 2007, 2010, 2011; Shapiro and Dinske, 2009a,b; Dinske and
Shapiro, 2013):
= lg N≥M − lg Q c (t) + bM. (5.21)
Equation (5.20) can be further rewritten in a more conventional form of the
Gutenberg–Richter law but with the a-value being time-dependent:
lg N≥M (t) ≈ at (t) − bM, (5.22)
with
at (t) = lg Q c (t) + . (5.23)
Thus, for fluid injections, the a-value becomes a function of the time that has
elapsed since the start of the injection. It is given by a sum of the seismogenic
index and the time-dependent cumulative volume of the injected fluid.
Finally, one more useful reformulation of equation (5.20) can be proposed. Let
us introduce a specific magnitude M defined as follows:

M = . (5.24)
b
Note that, as with the seismogenic index, the specific magnitude is completely
defined by seismotectonic features of the injection site. The larger the specific mag-
nitude, the larger the probability of significant induced events. Using the specific
magnitude, the Gutenberg–Richter law modified for fluid injections (5.20) can be
written in the following form:
lg N M ≈ lg Q c (t). (5.25)
Therefore, the number of events with M ≥ M is given just by the cumulative

volume of the injected fluid. Recently, Dinske and Shapiro (2013) observed this
type of behavior in real data. Note that both and M depend on the choice of the
metric system for the injected volume. We work in the System International and
measure Q c in m3 . Furthermore, we address moment magnitudes (see equation
(1.135)) in this book. One usually observes in the range from −8 to 1, and M
in the range from −4 to 1.
The statistical model of induced seismicity we summarized above can be equally
applied for non-monotonic injections. In such a case the probability (3.74) is
given by a minimum monotonic majorant of the pore pressure (see Parotidis
et al., 2004) and numerical computations are then required to predict lg N≥M (t)
for each particular situation.
The seismogenic index is a convenient quantity for a quantitative comparison
of seismotectonic activity at different locations. Here we compute the seismogenic
index at several borehole injection locations. Some of them are geothermal loca-
tions. Another part represents hydraulic fracturing of hydrocarbon reservoirs. For
our discussion we use time periods of the injections approximately corresponding
to non-decreasing fluid rates. We try to restrict our analysis to magnitude ranges
that are possibly weakly influenced by features of observation systems, of regis-
tration and of processing. Further, we attempt to use moment magnitudes. Where
moment magnitude are not given, our estimates of the seismogenic index will be
biased by the order of difference between the local magnitudes and the moment
magnitudes in corresponding magnitude ranges. At all locations shown here this
bias will be insignificant for our conclusions (see also comments below).
We consider familiar case studies. These are four geothermic locations: Cooper
Basin, Basel, Ogachi and Soultz. We also include non-geothermal sites: KTB (Jost
et al., 1998; Shapiro et al., 2006b), Paradox Valley (Ake et al., 2005), Cotton Valley
(Rutledge and Phillips, 2003) and Barnett Shale case study (Maxwell et al., 2006).
The last two locations correspond to hydraulic fracturing of gas reservoirs. The
German Continental Drilling Program Site (KTB) was a place of several borehole
fluid-injection experiments, the results of two of them corresponding to different
depths and conditions we show here.
Dinske and Shapiro (2013) estimated the seismogenic index for all loca-
tions mentioned above by using (5.21). The values of the seismogenic index are
shown in Figure 5.10. They demonstrate a reasonable stability (i.e. time inde-
pendence). They are stable regardless of the injection duration or the cumulative
injected fluid volume. Statistical errors and temporal fluctuations of the obtained
estimates are restricted in the interval of ±0.5. This is also the order of the bias
1
0 2
3
4
−2 5
6
7
8
−4
Σ
9
10
11
−6 12
13
14
−8 15
16
−10
0 0.2 0.4 0.6 0.8 1
Time/Injection Time
Figure 5.10 Seismogenic index computed for different locations of Enhanced

Geothermal Systems, hydraulic fracturing in hydrocarbon reservoirs, and other
injection locations (injection times are given in parentheses). 1: Ogachi 1991 (11
days), 2: Ogachi 1993 (16 days), 3: Cooper Basin 2003 (9 days), 4: Basel 2006
(5.5 days), 5: Paradox Valley (2500 days), 6-9: Soultz 1996 (48 hr), 1995 (11
days), 1993 (16 days) and 2000 (6 days), 10: KTB 2004/05 (194 days), 11–12:
KTB 1994 (9 hr) [upper and lower bound, calculated for two b-values], 13: Bar-
nett Shale (6 hr), 14–16: Cotton Valley Stages A (2.5 h), B (2.5 h) and C (3.5 h)
(Modified from Dinske and Shapiro, 2013.) A black and white version of this fig-
ure will appear in some formats. For the color version, please refer to the plate
section.
caused by differences between local and moment magnitudes. This fact is indirectly
confirmed by comparison of different magnitude studies for the same locations;
see, for example, publications of Mukuhira et al. (2008) and Deichmann and
Giardini (2009) on the Basel experiment.
We observe the following general tendency. The seismogenic index of hydraulic
fracturing sites is lower than the index of geothermal sites. In our examples of
hydrocarbon reservoirs it is lower than −4. It is larger than −3 for the geothermal
sites. Significantly higher has been estimated for the Cooper Basin data set. It is
of the order of −1.2 to −0.9. The largest index is obtained for the Basel data set:
0.1 to 0.7. The seismogenic index and b-value derived from seismicity in the sand-
stone and shale gas reservoirs in Cotton Valley and Barnett differ significantly from
corresponding quantities in geothermal reservoirs. The gas reservoirs are charac-
terized by a low seismogenic index, −10 < < −4, and a high b-value. The
rather large diversity of at Cotton Valley has the following reasons. In Stage
B (also called stage 2 in the previous chapter), the propagating hydraulic frac-
ture intersected and consequently opened a natural fracture system as shown in
Figure 4.6 (Dinske et al., 2010) that may have resulted in the highest value of
seismogenic index for the Cotton Valley reservoir. The fracturing stages were not
only performed in different boreholes but also with different treatment fluids: a
gel–proppant mixture was injected in Stage A (also called stage 3 in the previous
chapter) and B, whereas water was used in Stage C. Note that, for the fracturing in
Barnett shale, water was also injected. In both situations the value of is of the
same order.
For reservoir locations where multiple fluid injections had been carried out,
such as in Soultz or at the KTB, we observe several different indices. This is due
to the fact that the fluid was injected either in different wells and/or at different
depths. For example, at the KTB site (see Section 3.1) the 1994 injection was per-
formed at the depth of about 9 km using the main borehole to stimulate the SE1
fault zone, whereas in 2004/2005 the fluid was injected at the depth of ∼4 km
using the pilot borehole. In this latter experiment seismicity occurred along the
less-prominent fault zone SE2. Additionally, there was a long-term fluid-extraction
phase (one year) preceding the last injection. Also in Soultz the various injections
have stimulated different parts of the geothermal reservoir. A possible reason for
the differences in the derived seismogenic indices are different properties of the
fracture systems. Also multiple injections from the same borehole source can influ-
ence estimates of the index. As we have mentioned above, this can happen due to
reducing the bulk concentration of critically stressed defects (or equally, due to
elevating C max ) by preceding injections. In general, geothermal reservoirs and the
KTB as well as the Paradox Valley reservoirs are not only characterized by a higher
seismogenic index but also by a lower b-value (if compared to hydrocarbon case
studies). These observations lead to the conclusion that fluid injections in geother-
mal reservoirs have a higher potential to induce an earthquake having a significant
magnitude. This is in agreement with the observed earthquake magnitude ranges
(Majer et al., 2007).
Dinske and Shapiro (2013) analyzed possible correlations between the seismo-
genic indices and other parameters of injections, reservoirs and seismic hazard
estimates. In spite of the low data statistics due to the limited number of cata-
logs, they observed some interesting general tendencies. The seismogenic index
tends to be higher the greater the depth of a reservoir. They also observed a sig-
nificant positive correlation between the seismogenic index and the peak ground
acceleration, which is the seismic hazard indicator for naturally occurring seismic-
ity. Dinske and Shapiro (2013) found a correlative linear relationship between the
seismogenic index and the seismic hazard indicator specified as the peak ground
acceleration with 10% probability of exceedance in 50 years, corresponding to a
return period of 475 years (Giardini et al., 2003).
Summarizing, the seismogenic index helps us to characterize the level of seis-
mic activity one should expect by fluid injections into rocks. The index could
be estimated from a seismicity induced by a preliminary small-scale and short-
term injection test at sites where long-term and large-scale rock stimulations are
planned. Correlations between the seismogenic index and other parameters, such
as seismic hazard indicators, for example, can be also used to roughly estimate .
5.1.5 Occurrence probability of events with given magnitudes

A higher seismogenic index leads to a higher probability of significant events. Let
us first assume a constant injection rate Q I . Then, according to equations (5.5) and
(5.17), the cumulative event number at elapsed time t will be
N
Nev (t) = Q I t = ν I t, (5.26)
Cmax S
i.e. the cumulative event number is just proportional to the event occurrence time,
where ν I is a constant temporal event rate. Our model and assumptions described
above imply that the events occur independently from each other. In other words,
in our model the induced seismicity is assumed to be a Poisson process (like, for
example, a radioactive decay).
The Poissonian nature of seismicity processes can be tested by analyzing statistic
properties of waiting times between two successive events (the inter-event time).
Langenbruch et al. (2011) performed such an analysis. Here we briefly summa-
rize their main results. In a Poisson process with a constant expected event rate
ν I (the corresponding process is called a homogeneous Poisson process, HPP)
the probability Wn (ν I , t) of having n events in the time interval [0, t] is given by

(Hudson, 1964):
(ν I t)n
Wn (ν I , t) = exp(−ν I t). (5.27)
n!
The probability that, in a time interval t after a given event, no events will occur
is:
W0 (ν I , t) = exp(−ν I t). (5.28)
This probability can be interpreted as the probability W>t that the waiting time is
longer than t. The probability W<t that the waiting time is shorter than t is
then:
W<t = 1 − exp(−ν I t). (5.29)
Differentiating this expression in respect to t provides us with the probability
density function of this quantity, P D Ft :
P D Ft = ν I exp(−ν I t). (5.30)
Probability (5.29) can be further rewritten in the following form:
W<τ = 1 − exp(−τ ), (5.31)
where τ ≡ ν I t is the normalized waiting time. Therefore, we obtain the
following relation for the PDF of τ :
P D Fτ = exp(−τ ). (5.32)
Langenbruch et al. (2011) analyzed the inter-event time distributions of
injection-induced earthquakes for different catalogs collected at geothermal injec-
tion sites at Soultz and Basel. These catalogs are rather complete and contain large
numbers of events. They found that the distributions of waiting times during phases
of constant seismicity rate coincides quite well with the exponential distribution of
the homogeneous Poisson process (Figure 5.11).
Furthermore, Langenbruch et al. (2011) analyzed the waiting times for the com-
plete event catalogs and found that induced earthquakes are distributed according
to a non-homogeneous Poisson process in time. Moreover, they replaced the time
scale by the injected-volume one. They observed that in the fluid-volume scale the
seismicity process again becomes a homogeneous Poisson process. These results
strongly indicate that corresponding seismicity is directly induced by the fluid
injection.
Sequence of events with magnitude larger than M is also a Poisson process. The
corresponding event rate, ν M is equal to
ν M = ν I W≥M . (5.33)
100
10–1
10–2
PDF Δτ Soultz 93
Soultz 93
10–3 Soultz 94
Soultz 95
10–4 Soultz 96
Soultz 2000
Basel
10–5 HPP
exp(−Δτ)
10–6
10−2 10−1 100 101
Δτ
Figure 5.11 Estimated probability density functions of the normalized inter-event

time for stationary periods of injections at Soultz and Basel. (After Langenbruch
et al., 2011.) A black and white version of this figure will appear in some formats.
For the color version, please refer to the plate section.
Using the Poissonian distribution we obtain the occurrence probability Wn,M (t) of
n events of magnitude larger than M in the time interval (0, t):
(ν M t)n
Wn,M (t) = exp (−ν M t). (5.34)
n!
Let us now return to an arbitrary non-decreasing injection rate. Equation (5.10)
suggests that the induced seismicity is still a Poisson process but in respect to the
cumulative injected fluid mass rather than to the time. Moreover, it is a Poissonian
process in respect to the expected cumulative event number N≥M :
n
N≥M (t)
Wn,M (t) = exp −N≥M (t) . (5.35)
n!
In the case of a constant injection rate the process reduces to a standard homoge-
neous Poissonian process in time. Its distribution is given by equation (5.34).
Of practical importance is the probability of event absence. It is given by:

W0,M (t) = exp −N≥M (t) = exp −Q c (t)10−bM , (5.36)
where in the last part of the equation we have substituted the Gutenberg–Richter
magnitude distribution.
For an illustration we consider an injection experiment with parameters close to
the Basel case study. We assume that the seismogenic index is equal to 0.25. We
also assume that, during 5.5 days of the stimulation, 11 570 m3 of fluid have been
injected into the rock. We assume further that the b-value is approximately 1.5.
Equation (5.36) helps to answer the following question. What is the probability that
during this injection events with magnitude larger than M = 2.5 will occur? This
probability is given by substituting the injection parameters into (5.36), computing

the value of W0,M (t), computing the value of 1 − W0,M (t) and then multiplying the
result by 100%. We obtain approximately 97%. The probability that events with
magnitude larger than 3.5 will occur is approximately 11% (see also Figure 5.12).
Let us assume that we want to exclude occurrence of events with magnitude
larger than M with a probability W0,M (t) (e.g. 90%, i.e. W0,M (t) = 0.9). Then
(5.36) provides us with an estimate of a tolerable maximal cumulative injected
volume:
Q c (t) = −10bM− ln W0,M (t). (5.37)
Let us assume that we want to exclude the occurrence of events with magnitude
larger than M with a very high probability. This means that W0,M (t) is very close
to 1. In other words, δW0,M (t) = 1 − W0,M (t) is much smaller than 1. In this case
the previous equation can be expanded in a Taylor series and we approximately
obtain:
δW0,M (t) = Q c (t)10(−bM) . (5.38)
This equation provides the possibility of estimating the injection time, which is
allowed to exclude the occurrence of events with magnitude larger than M with a
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0
M
Figure 5.12 The probability that events with magnitude larger than a given one
would occur during a fluid injection. The parameters of the injection and of the
seismotectonic state are given in the text. They are close to those of the Basel
geothermal-stimulation experiment. (After Shapiro et al., 2010.)
given probability 1 − δW0,M (for instance, δW0,M = 0.001 means the absence of
non-desired events with a probability of 99.9%). Note that the condition of validity
of equation (5.38) is
Q c (t)10(−bM) 1. (5.39)
We recall that volumes are measured in m3 in all equations of this book.
If we assume a constant injection rate Q I , then Q c (t) = Q I t, and a require-
ment to exclude events with magnitude larger than a given one will reduce to the
restriction for the injection duration:
δW0,M (bM−)
t< 10 . (5.40)
QI
Therefore, the seismogenic index and the fluid injection rate are two key parameters
controlling the Poisson statistics of induced seismicity.
5.2 Statistics of large magnitudes

Fluid-induced seismicity results from stimulations of finite rock volumes. The
finiteness of stimulated volumes influences frequency–magnitude statistics. We
observe that fluid-induced large-magnitude events are frequently under-represented
in comparison with the Gutenberg–Richter law. This is an indication that the events
are more probable on rupture surfaces contained within the stimulated volume.
Following Shapiro et al. (2013) we will introduce different possible scenarios of
event triggering: rupture surfaces located completely within or intersecting only
the stimulated volume. These scenarios correspond to the lower and upper bounds
of the probability to induce a given-magnitude event. We will show that the bounds
depend strongly on the minimum principal axis of the stimulated volume. We will
further compare the bounds with the seismicity induced in several case studies.
The observed frequency–magnitude curves mainly follow the lower bound. Fitting
the lower bound to the frequency–magnitude distribution provides estimates of the
largest expected induced magnitude and the characteristic stress drop, in addition
to improved estimates of the Gutenberg–Richter a- and b-parameters.
5.2.1 Observations
From the previous sections we have seen that, during an active fluid injection
with non-decreasing injection pressure, the Gutenberg–Richter law modified for
injections (5.20) describes well the number N≥M of induced earthquakes with
magnitudes larger than M as a function of the injection time t.
Usually, large events have poor statistics (their number is usually less than 10),
which is not representative for a single-injection experiment. However, one can
identify systematic behavior by considering several injection experiments. Shapiro

et al. (2011) observed systematic deviations of large-magnitude statistics from
equation (5.20). The number of such events is significantly smaller than predicted,
especially for short injection-time intervals.
Let us first consider the geothermal fluid-injection experiment at Basel (Häring
et al., 2008; Dyer et al., 2008; Deichmann and Giardini, 2009). Figure 5.13 shows
observed and theoretical curves of N≥M (t). The simultaneous curve of the injected-
water volume Q c (t) is shown. Immediately after the injection termination, the
curve Q c (t) starts to decrease because of the outflow of the injected water. On a bi-
logarithmic scale, theoretical curves N≥M (t) corresponding to (5.20) are given by
a time-independent shifting of the curve of the injected-fluid volume downwards.
Note that (5.20) is applicable for injection periods only.
100 injected volume

M > 0.5 (Mc)
M>1
10–1 M > 1.5
M>2
10–2
10–3
100 101 102

Time since first event (h) shut-in time
Figure 5.13 Number N≥M (solid lines) of induced earthquakes with magnitudes
M larger than indicated values as functions of the time t elapsed from the time
of the first event in the catalog (nearly the injection start) at the Basel borehole
(Häring et al., 2008). The plot also shows the injected water volume Q c (t) (the
upper dashed line). The quantities Q c (t) and N≥0.5 (t) (the upper solid line; 0.5
is approximately a completeness magnitude) are normalized to their values at the
moment of the maximum injection pressure (several hours before the injection
termination). Immediately after the injection termination the curve Q c (t) starts to
decrease because of an outflow of the injected water. Theoretical curves N≥M (t)
corresponding to (5.20) are given by lower dashed lines. They are constructed by
a time-independent shifting of the curve of the injected-fluid volume. The lower
solid lines shows the observed quantities N≥M (t) normalized by the same value
as the quantity N≥0.5 (t). (Modified from Shapiro et al., 2011.) A black and white
version of this figure will appear in some formats. For the color version, please
refer to the plate section.
In a range of small magnitudes, the theoretical and observed curves are in good
agreement. However, the numbers of large-magnitude events are lower than the-
oretically predicted ones. This is seen from the fact that curves of N≥M (t) for
large magnitudes are no longer parallel to such curves for small magnitude ranges.
The large-magnitude curves deviate significantly downwards (especially for short
times elapsed after the injection start; note that corresponding theoretical predic-
tions give curves that are just parallel to the Q c (t) curve in the time period of the
injection).
We observe the same tendency at other locations. Figure 5.14 shows examples of
distributions N≥M as functions of the injected-fluid volume for several case stud-
ies. In all examples, an approximate agreement of equation (5.20) with numbers
for small-magnitude events has been demonstrated in the previous section. This
3
Cotton Valley (Gel) 3
Barnett Shale
10 10
M0 = −3.0
M0 = −2.2 M0 = −2.8
2 M0 = −2.0 2 M0 = −2.6
10 10
0
0
NM ≥M
NM ≥M
M0 = −1.8 M0 = −2.4
1 1
10 M0 = −1.6 10 M0 = −2.2
M0 = −1.4
0 0
10 10 4
2 3
10 10 102 103 10
3 3
VI [m ] VI [m ]
Soultz 1993 Basel

4 4
10 10
M0 = –0.8
M0 = 0.6
3 M0 = –0.4 3
10 10
M0 = 0.0 M0 = 1.0
0
0
NM ≥M
NM ≥M
2 2
10 M0 = 0.4 10 M0 = 1.4
M0 = 1.8
M0 = 0.8
101 101
M0 = 2.2
0 0
10 10
3 4
102 103 104 105 10 10
3
VI [m3] VI [m ]
Figure 5.14 Examples of N≥M as functions of the injected fluid volume at dif-
ferent sites. All curves correspond to the injection periods only. (After Shapiro
et al., 2013.)
is also seen in Figures 5.14. Indeed, according to equation (5.20) in a range of

not-too-large magnitudes, lines of lg N≥M (t) are nearly mutually parallel. Theoret-
ically, they all should be parallel to the function lg Q c (t). These lines are nearly
regularly spaced. Theoretically they should be equally spaced with an increment
given by the product of the b-value with the increment of the magnitude. However,
the number of large events is systematically smaller than a regular spacing of the
lines lg N≥M (t) would imply.
The fact that induced small-magnitude events obey the Gutenberg–Richter statis-
tics but large-magnitude events deviate from it indicates that the probability of an
earthquake on the corresponding rupture surface depends on the geometric relation
between this surface and the stimulated volume. Indeed, large-magnitude events
correspond to large ruptures. Such events are less common than small-magnitude
events. The statistics of rupture surfaces in rocks must correspond to the classical
Gutenberg–Richter statistics of earthquakes, which is a common observation. How-
ever, potential rupture surfaces of induced events (we consider a potential rupture
surface as the equivalent of a pre-existing crack) must intersect injection-stimulated
volumes. For a given rupture surface the probability to intersect a stimulated
volume will be higher the larger this rupture surface is. Thus, if we consider the pro-
portion of large potential rupture surfaces to small ones, we will observe that it is
significantly higher for the surfaces intersecting the stimulated volume than for rup-
ture surfaces in rocks in general. This suggests that large-magnitude events should
be over-represented in comparison to the Gutenberg–Richter statistics for small-
magnitude events. However, on the contrary we observe an under-representation of
large-magnitude events.
Thus, we will consider the two following extreme scenarios. In the first sce-
nario (for reasons that will become evident below, we call it the upper-bound
scenario, or just u-scenario) we assume that, to induce an event with a given rup-
ture surface, it is enough to stimulate an arbitrary small spot of this surface. In
induced seismicity, the portion of large-magnitude events in respect to the portion
of small-magnitude events should be higher than the proportion of large-scale
rupture surfaces in relation to small ones in rocks generally. Therefore, large-
magnitude events should be over-represented in comparison to the expectations
based on the Gutenberg–Richter statistics. This scenario seems to contradict the
observations.
In another scenario (we call it the lower-bound scenario, or just l-scenario) we
assume that, to induce an earthquake on a rupture surface, this surface must be
stimulated completely. This is in agreement with the following formulation of the
Mohr–Coulomb failure criterion: to enable an earthquake along a given interface,
an interface-integrated tangential stress must overcome the total friction force. As
soon as the largest part of a potential rupture surface remains unperturbed, the
Figure 5.15 A sketch of three different scenarios of inducing earthquakes on

given rupture surfaces. The ovals symbolize the stimulated volume. The light-gray
straight segments denote rupture surfaces of induced seismic events. The black
straight segments denote potential rupture surfaces, on which no earthquakes were
induced. The small circles on two light-gray segments in the middle sketch denote
rupture centers located within the stimulated volume.
probability of an earthquake remains low. Therefore, to enable an earthquake, a

significant part of the corresponding rupture surface (we assume: the complete
surface) should belong to the stimulated volume. This scenario seems to be more
adequate than the previous one.
Figure 5.15 schematically shows both the scenarios we discussed above. In addi-
tion, one more triggering situation is shown. The left-hand sketch corresponds to
the first considered scenario. Any intersection of a rupture surface with the stimu-
lated volume suffices for inducing the earthquake along this surface. The right-hand
sketch corresponds to the last considered scenario. An earthquake will occur only
if its rupture surface is located completely within the stimulated volume (it is the
only light-gray segment shown on the sketch). The figure shows one more situa-
tion. Earthquakes are induced on those rupture surfaces, whose centers (denoted in
the figure by the small circles on the corresponding segments) are located within
the stimulated volume. Note that the geometrical center of a rupture surface can
be replaced here by any characteristic point inside of the surface. Such a point can
be considered as a nucleation point of the earthquake. We can clearly see that the
number of the induced events decreases from the left to the right (from three to one
light-gray segment, respectively).
To quantify the scenarios shown in Figure 5.15 and to compare them with the
frequency-magnitude statistics of induced events we must modify the probability
W≥M from Section 5.1. The modification of this probability must take into account
the effect of the finiteness of the stimulated volume and of rupture surfaces. Note
that our scenarios are not directly related to the stimulation physics of potential
rupture surfaces. They just describe different possible statistical patterns of the
phenomenon of induced seismicity. Therefore, they are applicable to the seismicity
induced by elastic stress- or pore-pressure perturbations as well as to the seismicity
induced by other processes like rate- and state-dependent friction alterations. On
the other hand, a clear preference of the seismogenic process to follow one of these
scenarios can provide us with useful seismotectonic information.
5.2.2 Statistics of earthquakes with finite rupture surface

Let us assume that a finite volume V of rocks (the stimulated volume) has been
somehow altered so that seismogenic conditions in it have been changed suffi-
ciently to produce seismicity. In practice we assume that the stimulated volume is
approximately defined by an outer envelope of a cloud of hypocenters of induced
seismicity (see Figure 5.16). We consider the following simplified abstract model.
A stimulated volume is an ellipsoid or a cuboid that can grow with time (for
example, due to a fluid injection). Potential rupture surfaces (we identify them
with pre-existing cracks or faults) are randomly (or preferentially) oriented planar
circular discs (penny-shaped inclusions with vanishing thickness). The spatial dis-
tribution of centers of the discs is random and statistically homogeneous with the
bulk concentration N .
We introduce first the probability
W f (X ) = f X (X )d X, (5.41)
which is the probability of a given potential rupture surface in the unlimited

medium having diameter X . Note that f X (X ) is a PDF of a rupture surface of
the size X . For example, later, we will assume a power-law f X (X ) and show that it
1000
800
Soultz 1993
800 600
600 400
400 200
z [m]
200 0
North [m]
0 –200
–200 –400
–400 –600
–600 –800
–800 –1000
–500 0 500 –400 –200 0 200 400
East [m] x [m]
Figure 5.16 Two projections of the microseismic cloud of the case study Soultz
(1993) and of an approximating ellipsoid of the aspect ratio 43:33:10 (we use such
an ellipsoid for numerical simulations of geothermal stimulated volumes). The
left-hand-side projection is a map view. It corresponds approximately to the plane
of the intermediate and minimum principal axes (the L int , L min -plane). The right-
hand-side projection corresponds approximately to the plane of the maximum and
minimum principal axes (the L max , L min -plane). The minimum principal axis of
the ellipsoid is nearly parallel to the x-axis. (After Shapiro et al., 2011.)
corresponds to the Gutenberg–Richter statistics of earthquakes. Further, let Wc (X )

denote a probability that the center of the rupture surface (of diameter X ) belongs
to the stimulated volume under the condition that this surface intersects the vol-
ume. Let us also for an instant assume that all ruptures have the size X . It is clear
then that the product of Wc (X ) with the number of all rupture surfaces intersecting
the stimulated volume gives the number of rupture surfaces having their centers in
the volume. On the other hand, this number is equal to the product N V . Thus, the
ratio N V /Wc (X ) is the number of the all rupture surfaces intersecting the volume.
Recalling now that the rupture surfaces are statistically distributed over their size
we conclude that the ratio W f (X )/Wc (X ) gives the probability of a rupture surface
intersecting the stimulated volume having diameter X .
Now we can formulate the probability W E (X ) of an induced seismic event to
have a rupture surface of diameter X :
W E (X ) = Ws (X )W f (X )/Wc (X ) = G w (X )W f (X ), (5.42)
where in the second part on the right-hand side we have introduced one more nota-
tion: G w (X ) ≡ Ws (X )/Wc (X ). Ws denotes the probability of the corresponding
rupture surface being sufficiently stimulated to produce the event. This probability
is a conditional one. It implies that the rupture surface has something to do with
the stimulated volume. At least, the rupture surface intersects the volume (inter-
secting includes also touching). The presence of Wc (X ) in equation (5.42) shows
that there are many more rupture surfaces intersecting the stimulated volume than
just the product N V . Note that the left-hand sketch of Figure 5.15 shows in light-
gray all the rupture surfaces intersecting the stimulated volume. The middle sketch
shows in light gray the rupture surfaces corresponding exactly to the product N V .
In other words their number is equal to the product of the bulk concentration of
rupture surfaces (which is the bulk concentration of the centers of the potential
ruptures in rocks) with the stimulated volume V .
The quantity G w (X ) = Ws (X )/Wc (X ) describes the influence of the geometry
of the stimulated volume. G w takes also into account rupture surfaces intersecting
the stimulated volume. Depending on Ws , it can be larger or smaller than 1. Cor-
responding to the two scenarios considered in the previous section, this will lead
to W E > W f or W E < W f , respectively. Let us now concentrate on its part, the
probability Ws (X ).
Let us consider all potential rupture surfaces intersecting or located within the
stimulated volume. We recall that Wc (X ) is the probability of such a surface of
diameter X having its center within the stimulated volume. Let us also define a
probability of a seismic event along this whole rupture surface, We1 (X ), under the
condition that the rupture center is located within the stimulated volume but the
rupture surface is not entirely contained in the volume. Respectively, We2 (X ) will
denote the event probability for such a rupture surface under the condition that
its center is outside of the stimulated volume. We further compute the probability
Ws (X ) of a seismic event on a rupture of diameter X . Such an event occurs on a
rupture located completely within the stimulated volume. Its probability is given by
the product Wvol (X )Wc (X ), where Wvol (X ) is the probability for a rupture of size
X completely belonging to the stimulated volume under the condition that its cen-
ter is located within the volume. Such an event can also occur on a rupture with the
center inside the volume but intersecting the volume only. The corresponding prob-
ability is given by (1 − Wvol (X ))Wc (X )We1 (X ). Finally, such an event can occur
on a rupture intersecting the stimulated volume and having its center outside the
volume. The corresponding probability is given by (1−Wc (X ))We2 (X ). Therefore,
the probability of a seismic event along a rupture of diameter X is given by the sum:
Ws (X ) = Wvol (X )Wc (X ) + (1 − Wvol (X ))Wc (X )We1 (X )

+ (1 − Wc (X ))We2 (X ). (5.43)
In a general case considering any potential rupture surface intersecting the stim-
ulated volume (i.e. no pre-conditions for locations of rupture centers), the lower
bound Wsl of the probability Ws (X ) will be given by We1 (X ) = We2 (X ) = 0:
Wsl (X ) = Wc (X )Wvol (X ). (5.44)
This equation corresponds to the l-scenario (see the right-hand part of Figure 5.15).
For the upper bound of Ws (X ), Wsu (X ), several alternatives can be considered. The
largest and simplest one is We1 (X ) = We2 (X ) = 1, and thus,
Wsu (X ) = 1. (5.45)
This corresponds to the u-scenario discussed in the previous section (see the left-
hand part of Figure 5.15), where stimulation of an arbitrary small spot of a potential
rupture surface is enough for a corresponding seismic event. Equation (5.45) means
an over-representation of the large-magnitude events with respect to the standard
Gutenberg–Richter distribution. It can be seen from equation (5.42), where Wc (X )
becomes especially small for large X .
The next simple assumption would be We1 (X ) = 1 and We2 (X ) = 0. This
assumption means that, for triggering an event, the center of its potential rupture
surface must be within the stimulated volume (see the middle part of Figure 5.15).
Such a restriction is a reasonable formalization of the intuitive requirement that a
“significant part” or a “nucleation spot” of the rupture surface must be within the
stimulated volume (note also a topological equivalence between a disc center and
any other “nucleation center” placed inside the rupture). It leads to the following
estimate:
Wsu0 (X ) = Wc (X ). (5.46)
According to equation (5.42) this would mean that the statistics of induced
events should be given by W f (X ), i.e. given by the standard Gutenberg–Richter
distribution.
Three bounds (5.44)–(5.46) represent three different scenarios of the develop-
ment of induced seismicity. Their comparison with real data will clearly show
which of the scenarios is preferable for the induced seismogenesis.
The relation between the different estimates of Ws is
Wsl < Wsu0 < Wsu . (5.47)
Corresponding to the bounds of the probability Ws , we obtain the bounds for the
quantity G w = Ws /Wc :
G wl < G wu0 < G wu , (5.48)
where
G wl (X ) = Wvol (X ), (5.49)
G wu0 (X ) = 1, (5.50)
and, finally,
G wu (X ) = 1/Wc (X ). (5.51)
In the following section we discuss the probabilities Wc (X ) and Wvol (X )
associated with geometry of the stimulated volume.
5.2.3 Rupture-surface probability and geometry of stimulated volumes

Shapiro et al. (2011) investigated the probability Wvol (X ). This is the probability
that a disc of diameter X is completely contained within a given stimulated volume
under the condition that its center belongs to the volume. They found an exact
expression of the probability Wvol (X ) for a spherical stimulated volume of the
diameter L (see Appendix 1 in Section 5.3 of this chapter). In this case Wvol (X ) =
Wsp (X/L), where the function Wsp (X/L) is defined as follows:
2 1/2
X 1 X 2 X 3π X 3X X
Wsp ≡ 1+ 1− − + arcsin .
L 2 L L 4 L 2L L
(5.52)
The subscript “sp” indicates a spherical stimulated volume. This function is quickly
decreasing with increasing X/L. Therefore, Wvol (X ) can strongly influence the
statistics of induced earthquakes as soon as the size X approaches or exceeds a
100
num.Wsp
num.Wγ sp Soultz93
num.Wel Soultz93
anal.Wsp
probability (log10 scale)

10–1 anal.Wγ sp Soultz93
10–2
10–3
0 0.5 1 1.5 2
diameter/Lmin
Figure 5.17 Probability Wvol for a spherical and an ellipsoidal (geometry similar
to the case study Soultz 1993, see Figure 5.16) stimulated volumes shown as
functions of disc diameters normalized to minimum principal axes of the volumes.
The lines are given for spherical volumes (i.e. the exact functions Wsp ). The lower
line was computed for a spherical volume with a diameter equal to L min . The
upper line was computed for a spherical volume with diameter γ (see equation
(5.53)). The corresponding numerical results are given as circles and squares,
respectively. The crosses denote results of numerical modeling for the ellipsoid.
(After Shapiro et al., 2011.)
characteristic scale of the stimulated volume. We denote such a characteristic scale

as Y . It is clear that, in the case of a spherical stimulated volume, Y = L.
Shapiro et al. (2011) have numerically computed Wvol (X ) for ellipsoidal vol-
umes with principal axes L min < L int < L max . For numerical computations they
approximate a disc by a regular polygon with 16 sides (a hexadecagon). The cen-
ters of the polygons are distributed in a given stimulated volume with a given bulk
concentration. These polygons have random orientations. For a given X they com-
pute the number of polygons with all vertices located within the stimulated volume.
This serves as an approximate criterion that the discs are located completely within
the volume. Then, they normalize the result by the number of all centers. This gives
an approximation of Wvol (X ) (see Figure 5.17).
Shapiro et al. (2011) found that

1 −1/3
γ = 1/L min + 1/L int + 1/L max
3 3 3
(5.53)
3
usually provides a reasonably good estimate of a characteristic scale Y such that
Wvol (X ) becomes very small (Wvol (X ) 0.1) for X > Y . If L min is sufficiently
small, then it will provide a dominant contribution to γ . Thus, Y = O(γ ).
geothermal case hydrofrac case
100 100
Wel Wel
Wsph Wsph
numerical data numerical data
10–1 10–1
probability
probability
L min = 10
10–2 L min = 150 10–2 L inter = 50

L inter = 250 L max = 400
L max = 500 γ = 14.3842
γ = 201.2062
10–3 0 10–3
0.5 1 1.5 0 0.5 1 1.5 2 2.5 3
diameter / Lmin diameter / Lmin
Figure 5.18 A comparison of numerically computed (crosses) and theoretically

estimated (lines) probabilities Wvol as functions of disc diameters normalized
to the minimum principal axes of the volume. To the left: for an ellipsoidal
geothermal-type stimulated volume. The dashed line represents the approxima-
tion Wsp (X/γ ) (see the text below equation (5.53)). The solid line shows the
result of equation (5.87). To the right: the same, but for a hydraulic-fracture like
ellipsoid. The parameters of the ellipsoids and resulting size γ are given on the
plots. (After Shapiro et al., 2013.)
The function Wsp (X/γ ) will give a good approximation if the axes of the
stimulated ellipsoid are close to each other. Frequently (especially in the case of
hydraulic fracturing) one of the axes is extremely small: L min L int < L max .
In this case, further corrections are required to approximate Wvol (X ) (Shapiro
et al., 2013). Equation (5.87) from Appendix 2 (Section 5.4) defines a function
Wel (X ) providing such an approximation (see also Figure 5.18).
Shapiro et al. (2013) have also considered the situation with rupture inclined by
an angle ±φ to the plane of the maximum and intermediate tectonic stresses. From
Chapter 1 we know that this angle is defined by the friction coefficient and is usu-
ally close to ±30◦ (see Section 1.2). Thus, the rupture surfaces have a larger angle
to the minimum stress axis. To simplify the consideration, Shapiro et al. (2013)
assumed that the all potential rupture surfaces have these tilts. They approxi-
mated the stimulated volume by a cuboid with sides L min , L int and L max and
obtained:

X X X
Wvol (X ) = Wcub (X ) ≡ 1 − | sin φ| 1− | cos φ| 1− .
L min L max L int
(5.54)
For example, if φ = 0, all potential rupture surfaces will belong to the same
plane. Such geometry seems to be less relevant for the seismicity induced by fluid
stimulations of rocks. However, it is more adequate for aftershocks of earthquakes

in subduction zones.
The probability Wvol (X ) is important for the lower bound of the probability
W E (X ) of a seismic event with a rupture surface of the size X (see equations (5.42),
(5.44) and (5.49)). Another important probability, which can influence the statistics
of induced earthquakes is Wc (X ) (the upper bound of the probability W E especially
can be affected; see equations (5.46) and (5.51)) This is the probability of a rupture
surface of the size X having its center within the stimulated volume, given that
this rupture surface intersects the volume. Shapiro et al. (2013) investigated this
probability Wc (X ). Here we summarize their results.
Let us first consider chaotically oriented rupture surfaces and a spherical stim-
ulated volume of diameter L. The sought-after probability Wc (X ) (Figure 5.19)
is given by the ratio of the total number of rupture surfaces with the centers
within the stimulated volume to the total number of all rupture surfaces having
any intersections with (or completely located within) this volume (see Appendix 3
in Section 5.5):
−1
X 3 X 2 3π X
Wc (X ) = Wcsp ≡ 1+ + . (5.55)
L 2 L2 4 L
Figure 5.19 shows this result along with results of a numerical modeling.
100
numerical realization
analytical solution
NCin/NS
10−1
10−2
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
X/L
Figure 5.19 Probability Wc (X ) of a rupture surface of the diameter X having

its center within a spherical stimulated volume of the diameter L: analytical and
numerical results. (After Shapiro et al., 2013.)
Shapiro et al. (2013) have numerically investigated this probability for ellip-
soidal volumes with principal axes L min < L int < L max . Numerical results show
that substituting into equation (5.55) instead of L the following quantity
−2/3
1 3/2 3/2
γc = 1/L min + 1/L int + 1/L max
3/2
(5.56)
3
usually provides a good estimate of Wc . Note, again, that if L min is sufficiently

small, then it will provide a dominant contribution to γc .
For rupture surfaces tilted by an angle ±φ to a plane of the maximum and inter-
mediate tectonic stresses, Shapiro et al. (2013) approximate the stimulated volume
by a cuboid with sides L min , L int and L max , rather than by an ellipsoid. They
proposed the following approximation for Wc :
−1
X | sin φ| X | cos φ| X π X 2 | cos φ|
Wcc (X ) ≡ 1 + 1+ + + .
L min L max L int 4L max L min
(5.57)
Also here, if φ = 0 then all potential rupture surfaces belong to the same plane.
Equation (5.57) then provides probability Wc for a stimulated area of a rectangular
form. Such a geometry may be relevant for aftershocks of tectonic earthquakes.
5.2.4 Distributions of magnitudes and the Gutenberg–Richter law

We consider again the probability of an induced seismic event to have a rupture
surface of diameter X , W E (X ). This probability is formulated by equation (5.42).
Substituting equation (5.41) into (5.42) gives
W E (X ) = Ws (X )W f (X )/Wc (X ) = G w (X ) f X (X )d X. (5.58)
Therefore, the quantity G w (X ) f X (X ) is the probability density function of an

induced event with rupture surface of the size X . Now we must relate the prob-
ability density of magnitudes to the probability density G w (X ) f X (X ). Below, we
will closely follow the work of Shapiro et al. (2013).
The spatial scale of the rupture surface controls the magnitude of a correspond-
ing earthquake. A relationship between the rupture size X and the earthquake
magnitude M can be found by combining equations (1.133) and (1.135). This
yields the moment magnitude (Lay and Wallace, 1995; Shearer, 2009; Kanamori
and Brodsky, 2004)

M = 2 lg X + lg σ − lg C0 /1.5 − 6.07 (5.59)
for seismic moments measured in Nm. In the last part of the equation we con-
ventionally assume that the slip displacement scales as a characteristic length X
of the slipping surface (this is a result of the linear elastic theory of the fracture
mechanics; see also equation 9.26 and table 9.1 from Lay and Wallace, 1995, and
our equation (1.98)). The quantity σ is usually defined as a static stress drop,
and C0 is a geometric constant of the order of 1. We will use a shorter form of
equation (5.59):
M = 2 lg(X/Cσ ), (5.60)
where we have introduced a convenient notation Cσ for the cubic root of the
reciprocal stress drop: Cσ = 1084C0 /σ 1/3 ≈ 103 σ −1/3 .
1/3
In what follows we are interested in the statistic of magnitudes. Thus, we have to

consider M, X and Cσ as random variables. Equation (5.60) defines the magnitude
M as a function of two random variables, X and Cσ . It can be also written in the
following form:
X = Cσ 10 M/2 . (5.61)
This equation defines the rupture length X as a function of two random variables,
M and Cσ .
There are two transformation equations relating the pair of random variables
(M; Cσ ) to another pair (X ; Cσ ). The first relation, X (M, Cσ ), is given by equa-
tion (5.61). The second relation is given by the trivial statement: Cσ = Cσ .
These two relations define a coordinate transformation from the coordinate sys-
tem (M; Cσ ) to the system (X ; Cσ ). The Jacobian of this transformation is equal
to ∂ X (M, Cσ )/∂ M. This Jacobian and the transformation equation (5.61) yield the
PDF of magnitudes, f M :
f M (M) = 1.151 f X (Cσ 10 M/2 ) f C (Cσ )Cσ 10 M/2 , (5.62)
where we accepted ln 101/2 ≈ 1.151 and assumed that the random variables X and
Cσ are statistically independent. Furthermore, we introduced the following nota-
tions: f X (X ) is a PDF of the rupture length and f C (Cσ ) is a PDF of Cσ . Thus,
a probability W≥M of events with the magnitude larger than an arbitrary M is
equal to:
∞ ∞
W≥M = 1.151 f X (Cσ 10 M/2 ) f C (Cσ )Cσ 10 M/2 d MdCσ . (5.63)
0 M
Let us first assume that the following factorization is possible:
f X (Cσ 10 M/2 ) = f 1 (Cσ ) f 2 (10 M/2 ), (5.64)

where f 1 and f 2 are two independent functions. This will be the case if f X (X ) is a
power-law function. Then
∞
W≥M = AC f 2 (10 M/2 )10 M/2 d M (5.65)
M
with the proportionality coefficient

∞
AC = 1.151 f 1 (Cσ ) f C (Cσ )Cσ dCσ . (5.66)
0
Thus, under the factorizing assumption for f X , the randomness of the stress drop
influences the distribution of magnitudes by modifying its proportionality factor
(5.66) only.
Let us assume further a power-law PDF of a size of potential rupture surfaces
in an unlimited medium: f X (X ) ≈ A X X −β (here β > 0 and A X is a propor-
tionality constant). Note that such a PDF cannot be exactly valid because of a
possible integration singularity at X = 0. We assume that a power-law function
is a good approximation of a real PDF of potential rupture surfaces above a cer-
tain very small size (which corresponds to a magnitude significantly smaller than
M under considerations). Thus, a PDF f X (X ) strongly decreases with X . Power-
law size distributions are typical for natural fractal-like sets (Scholz, 2002; Shapiro
and Fayzullin, 1992; Shapiro, 1992). This type of self-similarity has been already
related to the Gutenberg–Richter frequency–magnitude distribution of earthquakes
(Shearer, 2009; Turcotte et al., 2007; Kanamori and Brodsky, 2004).
Indeed, a power-law PDF f X (X ) with β = 2b + 1 along with equations (5.60)
and (5.64) gives the following functions f 1 and f 2 :
f 1 (Cσ ) = A X Cσ−2b−1 , (5.67)
and
f 2 (10 M/2 ) = 10−bM 10−M/2 . (5.68)
Then equation (5.65) provides the Gutenberg–Richter law (we have previously
introduced it in the form of equation (5.11)):
W≥M = 10a−bM , (5.69)
where
∞
AX
a = lg f C (Cσ )Cσ−2b dCσ (5.70)
2b 0
and
b = (β − 1)/2. (5.71)
Therefore, a power-law size distribution of rupture surfaces leads to the

Gutenberg–Richter magnitude distribution (5.68) in a rather general case of an
arbitrary statistically distributed stress drop. In the following we will include in
our model a possibility of events of different size by assuming that the Gutenberg–
Richter magnitude distribution is the result of a power-law size distribution of
all rupture surfaces spanned by earthquake events. This assumption is supported
by the recent study of Langenbruch and Shapiro (2014). They showed that
Gutenberg–Richter relation results from the power-law size distribution of spa-
tial fluctuations of the Mohr–Coulomb failure stress caused by the fractal nature of
elastic heterogeneity of rocks.
5.2.5 Lower and upper bounds for magnitude distributions

In order to account for the finiteness of the stimulated volume we must include the
influence of the geometry into our considerations. Consequently, we must include
the quantity G w (see equation (5.42)) as a factor into the PDF of magnitudes. Tak-
ing into account (5.58) we see that we must replace f X (X ) by G w (X ) f X (X ) in
equation (5.63). Taking into account equation (5.64) and results (5.67)–(5.71) we
obtain:
∞ ∞
W≥M = 1.151A X 10−bM G w (Cσ 10 M/2 )Cσ−2b f C (Cσ )d MdCσ . (5.72)
0 M
Note that the quantity G w is usually a function of a ratio of the rupture scale X
and the characteristic scale Y of the stimulated volume. For example, in equations
(5.52)–(5.57) Y = L, γ , L min , respectively. Thus, the explicit dependence of G w
on Cσ can be eliminated by introducing a characteristic magnitude MY so that
Y = Cσ 10 MY /2 . (5.73)
Using this quantity, G w can be expressed (at least, approximately) in a modified

form G wm (M −MY ), which is directly obtained from G w by the corresponding sub-
stitution of the argument. For example, the most important for practical situations
is the lower bound of G w (X ). This bound is given by
G wl (X ) = Wvol (X ) ≈ Wsp (X/γ ). (5.74)
In turn,

Wsp (X/γ ) = Wsp 10(M−MY )/2 = G wm (M − MY ), (5.75)
where Y = γ .
Therefore, substituting the function G w (Cσ 10 M/2 ) by the function G wm (M −

MY ) in equation (5.72) we obtain:
∞
W≥M = 10ad 10−bM G wm (M − MY )d M, (5.76)
M
with the proportionality coefficient

∞
10ad = 1.151A X Cσ−2b f C (Cσ )dCσ = 2.303b10a . (5.77)
0
Generally, the magnitude MY is an unknown quantity effectively representing

the range of induced magnitudes. It is defined by the condition of equivalence of
equations (5.72) and (5.76). If Cσ−2b f C (Cσ ) tends to a narrow, δ-function-like dis-
tribution around a representative value Cσ then, in accordance with equation (5.61),
MY will be directly given by (5.73). In reality Cσ is restricted to a limited range
between approximately 1 and 1000. Fitting equation (5.76) to a real frequency–
magnitude distribution of an induced seismicity yields estimates not only of the
a- and b-values but also of the magnitude MY . By using equation (5.73) one can
estimate the representative value of Cσ and compute the corresponding estimate of
the stress drop. Note that, in the case of lower-bound magnitude probability (i.e.
G w = G wl ), owing to the vanishing probability Wvol for X > Y , the magnitude
MY is a limiting value of a largest possible magnitude of an induced earthquake.
From (5.59) and (5.73) we obtain:
MY = lg Y 2 + (lg σ − lg C0 )/1.5 − 6.07. (5.78)
Equation (5.76) can be also represented in the following form:
W≥M = 10a−bM+(M−MY ) , (5.79)
where
∞
−bm
(M − MY ) = lg 2.303b 10 G wm (m + M − MY )dm (5.80)
0
is a function correcting the magnitude distribution for the finiteness of the stim-
ulated volume (and m is an integration variable). This function can be roughly
estimated in the following way. The exponential function under the integral is a
quicker decreasing function than G wm (m + M − MY ). Thus, by integration we
can roughly assume that the last function is a constant equal to G wm (M − MY )
and we obtain (M − MY ) ≈ lg G wm (M − MY ). It clearly shows that, if M is
significantly smaller than MY (so that M − MY −1), the magnitude distribu-
tion will be indistinguishable from the classical Gutenberg–Richter one (because
G wm → 1). By M → MY the magnitude distribution will quickly drop down in
the case G wm → 0. This is the case for the lower bound of the function G w = G wl ,
which is given by equation (5.49).
We use (5.79) to further modify equation (5.20):
lg N≥M (t) = lg Q c (t) + + (M − MY ) − bM. (5.81)
Figure 5.20 shows theoretical cumulative frequency–magnitude curves (i.e. the
quantities lg N≥M as functions of M) for a given time that has elapsed since the
start of the injection. In the figure, the elapsed time has been involved implicitly
only. It defines corresponding geometrical sizes (L min , L int , L max ) reached by the
growing cloud of the seismicity. It also defines a value of the Gutenberg–Richter
quantity a(t) = lg Q c (t) + . For the particular examples shown in Figure 5.20,
a = 4.5 (e.g. = −0.5 and Q c = 105 m3 ) and b = 1.5. Shapiro et al. (2013)
numerically computed different functions (M − MY ) by using equation (5.80).
The functions G wm (m + M − MY ) were obtained by using substitution (5.61)
into the three functions G w (X ) given by equations (5.49)–(5.51), respectively. To
compute the lower bound of the quantity lg N≥M for the case of an ellipsoidal
stimulated volume, G w (X ) was substituted by the approximating function Wel (X )
defined by equation (5.87). To compute the lower bound of the quantity lg N≥M for
the case of a cuboidal stimulated volume, G w (X ) was substituted by the function
Wcub (X ) defined by (5.54). To compute the uppermost bound of lg N≥M for the
Lmin = 150.00 m Lint = 250.00 m Lmax = 400.00 m Lmin = 10.00 m Lint = 50.00 m Lmax = 400.00 m
7 7
Elllower bound Elllower bound
Ellupper bound Ellupper bound
6 6
Cublower bound Cublower bound
Cubupper bound Cubupper bound
5 5
Gutenberg–Richter Gutenberg–Richter
4 4
log10 Nev
log10 Nev
3 3
2 2
1 1
0 0
−1 0 1 2 3 4 5 −1 0 1 2 3 4 5
magnitude magnitude
Figure 5.20 Theoretical frequency–magnitude curves. From the lower to the

upper curves: the lower bound for the case of an ellipsoidal stimulated volume; the
lower bound for the case of a cuboidal stimulated volume; the Gutenberg–Richter
distribution; the uppermost bound for the case of a cuboidal stimulated volume;
the uppermost bound for the case of an ellipsoidal stimulated volume. On the left:
a geothermal-type of the stimulated volume. On the right: a hydraulic-fracturing
type of the stimulated volume. The scales of the stimulated volumes shown on
Figure 5.18. (After Shapiro et al., 2013.)
case of a cuboidal stimulated volume, the function G w (X ) was substituted by

1/Wcc (X ), which is reciprocal to the one given by equation (5.57). Finally, to
compute the uppermost bound of lg N≥M for the case of an ellipsoidal stimulated
volume, G w (X ) was substituted by the quantity 1/Wc (X ), which is reciprocal to
the approximative function given by equation (5.55) along with the quantity γc
from equation (5.56). Two situations are represented: geothermal- and hydraulic-
fracturing types of stimulated volumes. Both parts of the figure contain the five
following curves (from the lowest to the uppermost ones): a lower bound for an
ellipsoidal stimulated volume, a lower bound for a cuboidal stimulated volume, a
Gutenberg–Richter straight line, an upper bound for a cuboidal stimulated volume
and, finally, an upper bound for an ellipsoidal stimulated volume.
Note that the curves for ellipsoidal stimulated volumes are approximations only.
By contrast, the curves for cuboidal volumes are exact. However, the equations
for cuboidal stimulated volumes assume rupture surfaces inclined under an angle
φ to the plane of the maximum and intermediate axes. For Figure 5.20, Shapiro
et al. (2013) accepted φ = 30◦ . Figure 5.21 shows the curves for different values
of φ. Finally, Figure 5.22 shows an example of how a sophisticated geometric
form of the stimulated volume can influence the frequency–magnitude distribution
(a lower bound). Here a situation corresponding to two intersecting ellipsoids has
been numerically modeled.
Figures 5.20–5.22 and Figure 5.18 show that, if the seismicity statistics tends to
the lower bound, then a fitting of the Gutenberg–Richter straight line will produce
a systematically overestimated b-value. This effect will be stronger the smaller is
8 8
lower bounds lower bounds
upper bounds upper bounds
7 Φ = 0° 7 Φ = 0°
Φ = 5° Φ = 5°
6 Φ = 10° 6 Φ = 10°
Φ = 15° Φ = 15°
Φ = 20° Φ = 20°
5 Φ = 25° 5 Φ = 25°
Φ = 30° Φ = 30°
log10 Nev
log10 Nev
Φ = 35° Φ = 35°
4 Φ = 40° 4 Φ = 40°
Φ = 45° Φ = 45°
3 Gutenberg–Richter 3 Gutenberg–Richter
2 2
1 1
0 0
−2 −1 0 1 2 3 4 5 −2 −1 0 1 2 3 4 5
Magnitude Magnitude
Figure 5.21 The same as Figure 5.20 but cuboidal stimulated volumes and dif-
ferent angles φ. (After Shapiro et al., 2013.) A black and white version of this
figure will appear in some formats. For the color version, please refer to the plate
section.
107 single ellipsoid

two intersecting
106 ellipsoids
105
104
Nev
103
102
101
100
–2 –1.5 –1 –0.5 0 0.5 1
magnitude
Figure 5.22 Theoretical frequency–magnitude curves (lower bounds) for a stim-

ulated volume in a form of two intersecting ellipsoids.(After Shapiro et al., 2013.)
the size of the stimulated volume. The scale L min is especially important. Thus, the
effect will be especially strong for the hydraulic-fracturing type of the geometry.
This effect will also be strong for small time periods that elapse from an injection
start. For small injection times stimulated volumes are small. Thus, the effect will
result in a decrease of b-value estimates with injection times. This effect can be eas-
ily understood from equation (5.79). For the lower bound the quantity (M − MY )
is negative. Its absolute value is larger the smaller MY is and, therefore, the smaller
is the size of the stimulated volume. By fitting the Gutenberg–Richter straight line
this quantity will directly contribute to the values of the parameters b and a. It will
decrease a and increase b. This effect will act in the opposite direction, if the event
statistics follow the uppermost bound. It would increase a and decrease b values.
In the following section we compare several observed frequency–magnitude dis-
tributions to the theoretical bounds. The differences between the theoretical curves
for ellipsoidal and cuboidal volumes are not significant. Moreover, the angle φ is
not known. In spite of the fact that φ = 30◦ seems to be a reasonable approx-
imation, in reality the angle can be broadly distributed. Thus, following Shapiro
et al. (2013), we will compare real data to the theoretical approximations for
ellipsoidal volumes. In practice the stimulated volume can be satisfactory repre-
sented by an approximate outer ellipsoidal envelope of the cloud of hypocenters of
induced seismicity (see Figure 5.16).
5.2.6 Case studies of magnitude distributions and stress drop

Shapiro et al. (2013) compared equation (5.81) to frequency–magnitude
distributions in several case studies that are already familiar to us. These are two
geothermal locations in crystalline rocks, Basel (Häring et al., 2008) and Soultz
1993 (Baria et al., 1999). There is also the Paradox Valley data set obtained by
an injection of a saline water into deep carbonate rocks (Ake et al., 2005). Finally,
they have also considered three hydrocarbon locations: a hydraulic-fracturing Stage
A in a gas shale from Canada, a hydraulic fracturing Stage B (called stage 2
in the previous chapter) of a tight gas reservoir at the Cotton Valley (Rutledge
and Phillips, 2003), and a stage of a hydraulic fracturing from the Barnett Shale
(Maxwell et al., 2009).
The corresponding fitting results are shown in Figures 5.23–5.28. For their
analysis Shapiro et al. (2013) took microseismic clouds at the injection-termination
time t0 . In the first step they fitted the real data by a standard Gutenberg–Richter
cumulative distribution. They called the resulting straight line and its parameter
Soultz 1993 data

lower bound
4 Gutenberg−Richter GR
LB
3.5
2.5
log10 Nev
GR LB
b = 1.35 b = 1.06
2 a = 2.47 a = 2.85
1.5 Δσ = 382.782 Pa
MY = 1.31
1
MC = −0.71
0.5
0
−2 −1.5 −1 −0.5 0 0.5 1 1.5
magnitude
Figure 5.23 Fitting the frequency–magnitude distribution of the seismicity

induced by the Soultz 1993 injection. The axes of the ellipsoid representing the
stimulated volume are 440 m, 1400 m and 1740 m. The effective sphere scale γ
is approximately 630 m. Results of the fitting of the apparent Gutenberg–Richter
distribution (dashed line) and of the lower bound G wl (solid line) are shown on
the plot below the acronyms GR and LB, respectively. The thin upper straight line
shows the Gutenberg–Richter distribution with parameters a and b estimated from
the lower-bound curve. Mc , MY and σ are estimated values of the completeness
magnitude, the maximum induced magnitude and the stress drop, respectively.
HRB stage A data

lower bound
Gutenberg–Richter GR
3 LB
2.5
2
GR LB
log10 Nev
b = 1.33 b = 0.89
1.5 a = 0.90 a = 1.89
Δσ = 569.88 kPa
1 MY = 0.15
MC = −1.51
0.5
0
−2.5 −2 −1.5 −1 −0.5 0 0.5
magnitude

induced by a hydraulic fracturing stage A in a gas-shale deposit at Horn River
in Canada. Notations are explained in Figure 5.23. The axes of the ellipsoid are
10 m, 100 m and 650 m. The effective sphere scale γ is approximately 15 m.
apparent parameters of the Gutenberg–Richter distribution. Then they fitted the

data by the theoretical curve of the lower bound. Using the a- and b-values from
this fit they attempted to reconstruct “real” Gutenberg–Richter cumulative distribu-
tion. The reconstructed Gutenberg–Richter distribution describes seismicity in an
infinite medium with tectonic conditions of the corresponding injection site. The
re-estimated b-values are systematically lower than the parameters obtained by the
apparent Gutenberg–Richter fit. We observe that the real frequency–magnitude dis-
tributions are usually restricted between the lower bound and the fitting straight line
corresponding to the apparent Gutenberg–Richter distribution. Note that this line
is located lower than the magnitude–frequency distribution given by the function
G wu0 (see equation (5.50)) and corresponds to the reconstructed Gutenberg–
Richter distribution in the infinite medium. Moreover, nearly all data sets show
a tendency of the seismicity to be better represented by the lower bound.
The fitting of the lower bound also yields estimates of maximum induced mag-
nitudes and of an averaged stress drop. The stress drop is computed from equation
(5.78) by substituting the estimates of characteristic length γ and of the maximum
Barnett Shale data

lower bound
Gutenberg−Richter GR
LB
3
2.5
GR LB
log10 Nev
b = 2.72 b = 1.31
1.5 a = –5.02 a = –0.48
Δσ = 1.427 Pa
1 MY = –1.84
Mc = −2.73
0.5
0
–3.4 –3.2 –3 –2.8 –2.6 –2.4 –2.2 –2 –1.8
magnitude

induced by hydraulic fracturing at one of locations in Barnett Shale. Notations
are explained in Figure 5.23. The axes of the ellipsoid are 70 m, 340 m and
1000 m. The effective sphere scale γ was approximately 100 m. (After Shapiro
et al., 2013.)
induced magnitude MY . In reality the values of the stress drop of induced events
can be distributed in a broad range. For example, Goertz-Allmann et al. (2011)
estimated the stress drop of 1000 selected events from the Basel injection experi-
ment mentioned above. Their figure 2 shows values in the range of 0.1–102 MPa.
Jost et al. (1998) estimated stress drops of events induced by the 1994 KTB fluid
injection experiment. Their table 1 gives values distributed in the range 5 × 10−3 –
6 MPa. On the other hand, our lower-bound-based stress-drop estimate represents
an average value in the sense of the equivalence of equations (5.72) and (5.76). It
depends on a real distribution of the stress drop and will be dominated by the most
probable stress-drop values. Contributions of numerous small-magnitude events
can become enhanced. Such events can have rather small stress drops. However,
corresponding stress-drop values are frequently not seen because such events are
usually not analyzed owing to low signal–noise relations. The presence of such
events in the magnitude statistics can lead to even smaller averaged estimates of
3.5 Paradox Valley data

lower bound
LB
3
2.5
2 GR LB
log10 Nev
b = 0.83 b = 0.75
a = 3.59 a = 3.60
1.5
Δσ = 10.05 kPa
1
MY = 3.82
MC = 0.50
0.5
0
−1 0 1 2 3 4 5
magnitude

induced by the Paradox Valley injection. Notations are explained in Figure 5.23.
The axes of the ellipsoid are 3000 m, 4000 m and 7000 m. The effective sphere
scale γ was approximately 3800 m. (After Shapiro et al., 2013.)
stress drops than those measured at the KTB site. This tendency can be further
enhanced by the fact that the quantity γ is usually overestimated due to event loca-
tion errors. Still, because of its independence of any rupture model and any estimate
of the corner frequency, the lower-bound approach to estimating the stress drop can
yield a reasonable additional constraint of this poorly understood quantity.
The data sets permitting rather simple interpretation are shown in Figures
5.23–5.25. They correspond to a stimulation of a geothermal system in Soultz,
and to two hydraulic fracturing stages in the gas shale at Horn River (Canada) and
of Barnett Shale (USA).
The estimated values of the stress drop shown in the figures and containing many
digits were listed from the fitting algorithm directly. Realistically, these estimates
show only the order of magnitude of the stress drop. The Barnett Shale case study
(Figure 5.25) is distinguished by an especially low estimate of the stress drop. In
addition to the reasons already mentioned above, one more reason for this can
be the following. The minimum principal size of the stimulated volume in this
particular case study represents the total thickness of the shale reservoir. In reality
the induced seismicity is concentrated in several layers of the thickness of 10 m
Cotton Valley B data

lower bound
3 Gutenberg−Richter GR
LB
2.5
GR LB
log10 Nev
b = 1.89 b = 1.16
1.5 a = −0.86 a = 1.13
Δσ = 16.53 kPa
1
MY = –0.87
MC = –1.97
0.5
0
−2.5 −2 −1.5 −1 −0.5
magnitude

induced by the Stage B of hydraulic fracturing at Cotton Valley. Notations are
explained in Figure 5.23. The axes of the ellipsoid approximating the stimulated
volume are 10 m, 40 m and 480 m. The effective sphere scale γ is approximately
15 m. (After Shapiro et al., 2013.)
each (see Figure 4.13). Thus, in reality the quantity γ may be less by a factor of 10
than the one used for the estimate of the stress drop. This would yield a stress drop
of the order of 2500 Pa.
The complete data sets shown in Figures 5.23–5.25 are described well by
the lower-bound approximation. Thus the inducing of events seems to require
pore-pressure perturbations involving nearly total rupture surfaces. The fact that
the lower-bound curve somewhat underestimates the number of events in the
intermediate- to high-magnitude range can be explained by a too-rough analytical
approximation of the real rupture statistics. The influence of the geometry, which is
more complex than just an ellipsoid (see Figure 5.22), or a rather restricted angular
spectrum of the rupture orientations (see Figure 5.21) could also contribute to this
effect.
A somewhat more-sophisticated interpretation seems to be required by the data
sets shown in Figures 5.26–5.28. These are data from the Paradox Valley, Cotton
Valley (Stage B) and Basel case studies.
Basel data
lower bound
3.5
LB
2.5
GR LB
log10 Nev
2 b = 1.44 b = 1.33
a = 4.28 a = 4.32
1.5
Δσ = 12.46 MPa
MY = 3.05
1
MC = 0.55
0.5
0
0 0.5 1 1.5 2 2.5 3 3.5
magnitude

induced by the fluid injection at the Basel EGS site. Notations are explained in
Figure 5.23. The axes of the stimulated volume are approximately 100 m, 760 m
and 920 m. The effective sphere scale γ is approximately 150 m. (After Shapiro
et al., 2013.)
In Figures 5.26 and 5.27 we again observe that a dominant majority of events
follow the lower-bound approximation well. However, several data points cor-
responding to high-magnitude events tend to return backward to the classical
Gutenberg–Richter distribution. Thus apparently, the corresponding few large-
magnitude events were triggered merely by a pore-pressure-related perturbation
of nucleation spots on their rupture surfaces. Thus, possibly the rupture surfaces of
these events were not completely included in the stimulated volume. However, for
their triggering an excitation of rather large nucleation domains was necessary. We
conclude this from the fact that corresponding data points are still located below
the Gutenberg–Richter distribution. This indicates that excitation of just a nucle-
ation spot was not sufficient for their occurrence. Note also that we cannot exclude
the possibility that these few large-magnitude events correspond to rare statistical
fluctuations with high stress drops and/or spatial orientations permitting large-scale
(significantly larger than γ ) rupture surfaces.
One more example of a similar situation is given by the Basel data set (Figure
5.28). A conventional Gutenberg–Richter fitting yields a = 4.3 and b = 1.4.
The bound G wl yields close results: a = 4.3 and b = 1.3. Additionally, σ =

12.5 MPa. The maximum magnitude defining the lower bound is found to be MY =
3.05. It seems that, for the triggering of a majority of events in Basel, stimulation
of their nucleation spots was sufficient.
5.2.7 Induced and triggered events

The previous discussion of case studies shows that magnitude distributions on
Figures 5.23–5.25 are described well by the lower-bound approximation. The situ-
ations shown in Figures 5.26–5.28 seem to be more complicated. A comparison
of these two groups of case studies indicates the possibility of distinguishing
between triggered and induced events. McGarr et al. (2002) defined induced and
triggered seismicity in respect to the stress impact of a stimulation. If this impact
is of the order of the ambient shear stress they speak about induced seismicity.
If this impact is significantly smaller than the ambient shear stress they speak
about triggered seismicity. In this terminology, nearly all the fluid-induced seis-
micity from our case studies could be considered as the triggered ones. Here
we use another definition, which is more specific in respect to the geometry of
the stimulated volume. It is close to the one used by Dahm et al. (2013). We
define as “induced” those events resulting from perturbing their nearly complete
rupture surfaces. Then their statistics should follow the lower bound of magni-
tude distributions. We define as “triggered” those events resulting from perturbing
nucleation spots of their rupture surfaces only. Their statistic should follow the
reconstructed Gutenberg–Richter distribution. It seems that some geothermal-
reservoir case studies include triggered events (e.g. the Basel case study). This
is rare but cannot be excluded for hydraulic fracturing of hydrocarbon reservoirs.
Note also that, for an event triggering, a perturbation of an arbitrary element
of its rupture surface is not sufficient. This would correspond to the uppermost
bound (5.51). This bound strictly contradicts our observations. The data show
that triggering requires a perturbation of a significant part of the rupture surface
necessarily including the nucleation domain (which we model by the rupture cen-
ter). It also seems that the induced events are much more common than triggered
events.
5.2.8 Maximum magnitude and sizes of stimulated volume

In Section 5.2.3 we saw that the probability Wvol (X ) quickly decreases with
increasing size X of the rupture surface. As soon as the size X approaches or
exceeds a characteristic scale Y of the stimulated volume the probability of events
becomes nearly vanishing. Our numerical simulations of Wvol show that the size
of the stimulated volume defines the scale Y . In comparison to other geometric

parameters, the length L min gives a dominant contribution to the magnitude order
of Y (see the quantity γ given by equation (5.53)).
The conclusion on a dominant role of L min in the parameter Y is supported by
real data. Shapiro et al. (2011) considered the largest observed magnitudes Mmax of
induced earthquakes for different case studies. They assumed Mmax to be a proxy
of MY and showed that the values of Mmax are much better described as functions
of lgL 2min than as functions of lgL int
2
or lgL 2max (see their figure 7).
Equation (5.81) shows that the frequency–magnitude statistic of induced earth-
quakes is time dependent. This is not only due to the term containing Q c (t).
Another time-dependent factor is the characteristic magnitude MY . This magni-
tude is a function of a geometrical scale of the stimulated volume. This scale is, in
turn, a function of time (or, equivalently, of the injected volume). This means that
both quantities a and b of a conventionally fitted Gutenberg–Richter frequency–
magnitude distribution will be time dependent. We have already mentioned that one
should expect a systematic increase of the a-values and decrease of the b-values
with increasing time that has elapsed since the start of the injection. However, infor-
mation on large-magnitude events at intermediate time moments can be included
into a consideration of the maximum observed magnitudes as a function of the
scale of stimulated volumes.
Figure 5.29 (reproduced from Shapiro et al., 2013) shows the observed maxi-
mum magnitudes as function of the minimum axis of the stimulated volume. The
figure includes maximum magnitudes for the injection termination moment t0 as
well as maximum magnitudes for the time moments t2/3 defined so, that the injected
volume Q c (t2/3 ) numerically satisfies the following condition: lg Q c (t2/3 ) =
(2/3) lg Q c (t0 ) (all volumes are measured in m3 ). We see that the data support
the general trend shown by the straight line. The error bars on the figure show
the possible impact of errors in magnitudes and event locations. For magnitudes,
Shapiro et al. (2013) assumed an error of the order 0.5. This roughly corresponds to
possible differences between local and moment magnitudes (see a thorough study
of this subject by Grünthal and Wahlström, 2003), which were taken from dif-
ferent literature sources (see the citations below). For the principal axes, Shapiro
et al. (2013) assumed the error bars of the order of seismicity location errors:
10 m for hydraulic-fracturing sites Barnet Shale (Maxwell et al., 2009) and Cot-
ton Valley (Rutledge and Phillips, 2003), 50 m for geothermic sites: Basel (Häring
et al., 2008), four experiments of Soultz (the data courtesy of Andrew Jupe, EGS
Energy), two experiments of Cooper Basin (Baisch et al., 2009), Fenton Hill
(Phillips et al., 1997), Berlin (Bommer et al., 2006), and 100 m for the Paradox
Valley (Ake et al., 2005). The star corresponds to the largest event of the Basel
injection.
2
Mmax
–1
–2
–3
–4
0 1 2 3 4 5 6 7 8 9 10
lg L2
min
Figure 5.29 Largest observed magnitudes of induced earthquakes as a function

of the minimum principal axes of corresponding stimulated rock volumes for dif-
ferent case studies. The length of L min is given in meters. The error bars show
the possible impact of errors in magnitudes and event locations. The star cor-
responds to the largest event of the Basel injection. The crosses correspond to
various data sets including also seismicity clouds at the time moments t2/3 such
that the cumulative injected volume Q c (t2/3 ) at these times numerically satisfies
lg Q c (t2/3 ) = (2/3) lg Q c (t0 ), where t0 is the termination time of the stimula-
tion and all volumes are measured in m3 . Injection sites, 2 lg L min and maximum
magnitudes for the data catalogs taken till the moment t0 of injection termination
(the latter two quantities are given as a coordinate pair in the coordinate system of
the figure) are: Barnett Shale ( 3.69; −2.0), Cotton Valley A (2.6; −1.2), Cotton
Valley C (2.6; −1.7), Basel (4.83; 3.05), Paradox Valley (6.95; 4.3), Soultz 1993
(5.29; 1.0), Soultz 1995 (4.89; 0.1), Soultz 1996 (4.98; 0.3), Soultz 2000 (5.4;
2.3), Berlin (Bommer et al., 2006) (6.23; 2.0), Cooper Basin 2003 (5.59; 3.7),
Cooper Basin (4.95; 2.9) and Fenton Hill (5.2; 1.3). (After Shapiro et al., 2013.)
We observe good agreement of the data points with equation (5.78) for Y =
L min . The corresponding values of σ are of the order of 10−4 –10 MPa. Substi-
tuting a highest probable limit of stress drops of the order of 10 MPa into equation
(5.78) we obtain an approximate estimate of the maximum probable magnitude
limit (it would correspond to an upper envelope on the Figure 5.29) of an induced
earthquake for a given location:
max{Mmax } ∼ 2 · lg L min − 1. (5.82)

5.3 Appendix 1. Probability of discs within a sphere 253
This result explains the fact that high-magnitude events are more probable at
geothermal sites than via hydraulic fracturing in hydrocarbon reservoirs. L min is
much smaller in the latter case than in the former one. By hydraulic fracturing,
a typical zone of water penetration behind the fracture walls is of the order of
1–10 m (owing to localization errors, microseismic clouds of hydraulic fractures
can have greater L min ; examples of a high-precision localization of hydraulic-
fracture induced microseismicity can be found in Rutledge and Phillips, 2003;
some of them have been shown in the previous chapter). Thus, the largest mag-
nitudes are 0–1. By geothermal stimulations, L min is of the order of 300 m or less.
In such a case the largest magnitudes are around 4. Because L min can increase with
time, it leads to an enhancement of the large-magnitude probability at the end of
injection operation or shortly after it (until the rate of induced seismicity has not
dropped significantly). This has been indeed observed in reality (Majer et al., 2007;
Baisch et al., 2009).
Result (5.82) can be useful for estimating and constraining an induced seismic
hazard. For example, to restrict the hazard one could attempt to keep the mini-
mal principal axis of the stimulated volume restricted by terminating the injection
if this size achieves a planned critical value. However, because this result mainly
addresses induced events, its application requires a careful analysis of the seis-
motectonic and geological situation in each particular case. By such an analysis,
the probability of triggered events should be carefully constrained. Factors such
as scales of faults intersecting the stimulated volume, tendencies of the faults to
influence the shape and scales of the stimulated volume, and other seismotectonic
parameters (e.g. the seismogenic index, b-values, stress states of significant faults
and modifications of these parameters during the stimulation) are of importance for
this task.
The formalism of the bounds of magnitude probability is not restricted to fluid-
induced seismicity. We hypothesize that it is applicable for any type of seismicity
induced in a restricted rock volume, for example an aftershock series of tectonic
events.
5.3 Appendix 1. Probability of discs within a sphere

Here we consider the probability of an arbitrary oriented disc, which has its center
inside a spherical volume, to completely belong to this volume. As a result we will
derive equation (5.52). We follow the derivation of Shapiro et al. (2011).
Let us first find possible locations of centers P of such discs. We consider a
spherical stimulated volume of a radius R with a center at a point O. We take a disc
with a radius r < R and a center at a point P. P is located inside of the stimulated
volume. If the distance y = |O P| is less than R − r , then the volume will include
the disc independently of its orientation. Thus, the volume V1 = 4π(R −r )3 /3 will
contribute to a possible location of P completely. Let us consider the spherical
√ shell
with the inner radius equal to ymin = R − r and the outer radius ymax = R − r 2 .
2
The discs belonging to the volume cannot have centers P outside the outer sphere
of the radius ymax .
If P is located inside the shell then the disc cannot be arbitrary oriented. An
orientation of the disc is given by the orientation of a normal to its plane at the
point P. Now we must consider possible orientations of the discs. Let us consider
a sphere S1 of radius r with the center at P. This sphere intersects the surface of
the stimulated volume along a circle. We consider further a plane including the
straight line OP. Such a plane intersects this circle in two points A1 and A2 located
symmetrically to the line OP. The sine of angle α between the line PA1 (or PA2)
and the normal to the line OP at the point P belonging to the same plane (given by
points O,P, A1 and A2) can be computed as sin α = (R 2 − r 2 )/(2yr ) − y/(2r ).
In order to belong to the volume a disc must have its normal located inside the
cone with the symmetry axes OP and the limiting angle α. The sphere S1 and this
cone define a spherical segment of the height h = r (1 − cos α) and the surface
2πr h. The probability of a disc having an orientation necessary to belong to the
stimulated volume is equal to the ratio of this surface to the surface of the half of
the sphere S1, i.e. h/r . This probability W1 is a function of the variable y.
The probability we are looking for is given by the sum of the contribution of
the volume V1 and of the integral of the probability W1 over the spherical shell
introduced above:
ymax
2h
Wsp (r/R) = 3 y dy + (R − r ) /R 3 ,
3
(5.83)
ymin r
√
with ymin = R − r and ymax = R 2 − r 2 . This provides equation (5.52).
5.4 Appendix 2. Probability of discs within an ellipsoid

Here we consider the probability that an arbitrary oriented disc with the center
inside an ellipsoidal stimulated volume completely belongs to the volume. We fol-
low the corresponding derivation of Shapiro et al. (2013). In contrast to the exact
results (5.52) and (5.54), we propose an approximation of the required probability
Wel (X ) (where X is the diameter of the rupture). We consider an ellipsoid with
the principal axes L min < L int < L max . If the axes are close to each other, then
a good approximation of Wel (X ) will be Wsp (X/γ ), where γ is given by equation
(5.53). Let us consider another quite a realistic geometry of stimulated volumes:
L min L int < L max .
5.4 Appendix 2. Probability of discs within an ellipsoid 255
We consider a rupture surface in a form of a plane disc of diameter X of an

arbitrary orientation with a center at a point P inside such an ellipsoidal stimulated
volume. We will concentrate first on large discs with X > L min . The centers of
large discs completely belonging to the ellipsoid are approximately located inside
the following ellipsoidal volume:
π
Vlarge = L min (L int − X )(L max − X ). (5.84)
6
We consider a sphere S1 defined by normals of the length X/2 (the sphere’s
radius) at the point P for all possible orientations of the disc. If X > L min (i.e.
large discs) then such a sphere will always intersect the surface of the ellipsoid.
We estimate approximately a part of the surface of this sphere (in relation to the
spheres complete surface), where the disc’s normal can have its end point for a disc
belonging to the stimulated volume. For this we consider such a sphere intersecting
with a “side surfaces” of the volume. The side surfaces are two ellipsoidal surface
halves spanned on the axes L int and L max .
Further, we approximate these surfaces just by planes separated by L min (we
call them side planes). We consider the sphere S1 of radius X/2 with the center P
at a distance y from the closest side plane. This sphere intersects the side plane
along a circle. In order to belong to the volume, a disc must have a normal located
inside a cone with the symmetry axes coinciding with the the normal from P to the
side plane. The sphere S1 and this cone define a spherical segment of the height
h = (X − X 2 − 4y 2 )/2 and the surface π X h. A probability of a disc having an
orientation necessary for belonging to the stimulated volume isequal to the ratio of
this surface to the surface of the half of the sphere S1, i.e. 1 − 1 − 4y 2 / X 2 . Note
that the effect of a possible intersecting of the sphere S1 with the second (farthest)
side plane is automatically taken into account. Indeed, the largest intersection is
of importance only because of its symmetric effect on permitted orientations of
the discs. The probability W L of large discs belonging to the stimulated volume is
given by the integration over y:

2(L int − X )(L max − X ) L min /2 4y 2
W L (X ) = 1 − 1 − 2 dy. (5.85)
L min L int L max 0 X
The integration yields:

⎛ ⎞
2
⎝ 1 L min X X ⎠ X X
W L (X ) = 1 − 1− 2 − arcsin 1− 1− .
2 X 2L min L min L int L max
(5.86)
From the derivation it is clear that, with increasing X ≤ L int , the estimate
W L (X ) will adequately decrease to zero. However, for small X the function W L (X )
becomes inadequate. At the point X = L min it must be reasonably combined with

the function Wsp (X/γ ). Thus, we propose the following approximation of Wel (X ):
Wel (X ) ≈ Wsp (X/γ ), if X ≤ L min ,

Wsp (L min /γ )
Wel (X ) ≈ W L (X ) , if L min ≤ X ≤ L int . (5.87)
W L (L min )
5.5 Appendix 3. Probability of discs with centers inside a sphere

Here we consider the probability that an arbitrarily-oriented disc that intersects a
sphere has its center within this sphere. We follow the corresponding derivation of
Shapiro et al. (2013).
We consider a spherical volume (a stimulated volume) of the radius R = L/2
with a center at the point O (see Figure 5.30). To calculate a number Ns of discs of
the radius r = X/2 intersecting this volume, we must locate their centers (points
P) inside or outside the stimulated volume. Such discs can belong to two different
groups.
The first group consists of discs with centers P belonging to the stimulated vol-
ume. Their number is is given by 4π N R 3 /3, where N is the bulk concentration of
the discs (i.e. the bulk concentration of the potential rupture surfaces).
A1
A A2
O O1
A4
P A3
A1
A2
A
O O1 A4
o2 P A3
Figure 5.30 Geometrical sketches for computing probability of an arbitrarily ori-

ented disc intersecting a spherical volume having its center inside the volume.
5.5 Appendix 3. Probability of discs with centers inside a sphere 257
The second group consists of discs with centers outside the stimulated volume.
Let us consider a sphere S1 of radius r with the center at P. An orientation of the
disc is given by the orientation of the normal to the disc at the point P. A necessary
condition for the disc to intersect (or to touch) the stimulated volume is that the
sphere S1 completely includes the stimulated volume or intersects the surface of
the stimulated volume along a circle (or just touch the volume). We introduce a
variable s = |O1P| = |O P| − R (see Figure 5.30) and consider further a plane
including the straight line OP. In this plane there are exist two radii of S1 which
(or continuations of which) are tangential to the spherical stimulated volume. We
denote one of the touching points as A (see the top part of Figure 5.30). There are
two such points located
symmetrically to the line OP. From the triangle OAP we
have |A P| = τ = (R + s)2 − R 2 . We consider then two possible situations: (1)
τ ≤ r (shown in Figure 5.30, on the top), and (2) τ > r (shown in Figure 5.30 on
the bottom; note that here A denotes an intersection point of the surface of S1 and
of the surface of the stimulated volume).
In case (1) the limiting orientations of a disc intersecting the volume are given
by the tangential positions of the radii of S1 (Figure 5.30, on the√top). Note that
this yields the following restrictions for the variable s: 0 ≤ s ≤ r 2 + R 2 − R.
A disc will intersect the stimulated volume if the disc’s normal is located within
the spherical sector defined by rotation of the plane section A1PA2 around the
line PA4. Therefore, such a normal can intersect the surface of the right-hand
(upper) half-sphere of S1 everywhere excluding the surface of the spherical seg-
ment covered by the rotation of the semi-arc A2A4 around the point A4. Taking
into account that sin O P A = sin A1P A2 = sin P A2A3 = R/(R + s) we obtain
|A3A4| = r s/(R + s). For the segment surface and the sector surface this yields
2πr 2 s/(R + s) and 2πr 2 R/(R + s), respectively. The probability of a disc having
an intersection with the volume is given by the relation of the spherical sector sur-
face to the surface of the half sphere. This ratio is equal to R/(R + S). The number
of discs intersecting the stimulated volume is then given by the following integral:
√
r 2 +R 2 −R
4π N (R + s)Rds = 2π Rr 2 N . (5.88)
0
An additional number of rupture surfaces intersecting the volume is given by

√ (2): τ > r . This situation corresponds to the bottom part of Figure 5.30. Here
case
r 2 + R 2 − R ≤ s ≤ r . The straight line PA is no longer a tangent to the stimulated
volume. The height of the excluded spherical segment (defined by the rotation of
the semi-arc A2A4) is given by r −|P A3|. The length h 1 = |P A3| = |AO2| can be
found from the triangle OAP. The probability of a disc intersecting the stimulated
volume is given by

h 1 /r = 2(R + s)2 (R 2 + r 2 ) − (R + s)4 − (R 2 − r 2 )2 /(2r (R + s)). (5.89)
The number of rupture surfaces intersecting the stimulated volume is then given by
the following integral:
r
4π N √ (R + s)2 (h 1 /r )ds = π 2 R 2r N . (5.90)
r 2 +R 2 −R
Finally, the sought-after probability Wc (X ) is given by the following ratio, taking

into account all the contributions discussed above:
−1
4π N R 3 /3 3 X 2 3π X
Wc (X ) = 2 2 = 1+ + .
π R r N + 2π Rr 2 N + 4π N R 3 /3 2 L2 4 L
(5.91)
5.6 Appendix 4. Probability notations used in this chapter
symbol its meaning

Wn (ν I , t) probability of n events in the time interval [0, t]; ν I is the event rate
W>t,<t probability of waiting time being longer (shorter) than t
Wn,M (t) probability of n events of magnitude not less than M in the time interval
[0, t]
Wev probability of a point-like event
W≥M probability of an earthquake with a magnitude not less than M
W E (X ) probability of an event with rupture-surface diameter X
W f (X ) probability of a potential rupture surface with diameter X , in unbounded
medium
Wc (X ) probability of a rupture surface with the center within the stimulated volume
Ws (X ) probability of a rupture surface being sufficiently stimulated to produce an
event
Wvol (X ) probability of a rupture surface being completely inside the stimulated
volume
We1 (X ) probability of a seismic event along a given rupture surface under the
condition that the center of the surface is inside of the stimulated volume
We2 (X ) probability of a seismic event along a given rupture surface under the
condition that the surface is intersecting the stimulated volume but its center
is outside of the volume
Wsl (X ) lower bound of Ws (X ): rupture surface is inside the stimulated volume
Wsu (X ) upper bound of Ws (X ): arbitrary intersection of the rupture surface and
stimulated volume triggers the event
Wsu0 (X ) upper bound of Ws (X ): center of the rupture surface has to be inside the
stimulated volume to trigger the event
Wsp Wvol for a sphere
Wel approximation of Wvol for an ellipsoid
Wcub Wvol for a cuboid
Wcc Wc for a cuboid
References
Ake, J., Mahrer, K., OConnell, D., and Block, L. 2005. Deep-injection and closely mon-
itored induced seismicity at Paradox Valley, Colorado. Bulletin of the Seismological
Society of America, 95, 664–683.
Aki, K., and Richards, P. G. 2002. Quantitative Seismology, 2nd edn. Sausalito, CA:
University Science Books.
Al-Wardy, W., and Zimmerman, R. W. 2004. Effective stress law for the permeability
of clay-rich sandstones. Journal of Geophysical Research: Solid Earth, 109(B4),
doi:10.1029/2003JB002836.
Altmann, J. B., Müller, T. M., Müller, B. I. R., Tingay, M. R. P., and Heidbach, O. 2010.
Poroelastic contribution to the reservoir stress path. International Journal of Rock
Mechanics and Mining Science, 47, 1104–1113.
Amenzade, Yu. A. 1976. Theory of Elasticity. Moscow: Visshaja Shkola, pp. 210–211 (in
Russian).
Auld, B. A. 1990. Acoustic Fields and Waves in Solids, 2nd edn, vol. 1. Malabar, Florida:
R. E. Krieger.
Avouac, J.-P. 2012. Earthquakes: human-induced shaking. Nature Geoscience, 5, 763–764,
doi:10.1038/ngeo1609.
Backus, G., and Mulcahy, M. 1976a. Moment tensors and other phenomenologi-
cal descriptions of seismic sources I. Continuous Displacements. Geophysical
Journal of the Royal Astronomical Society, 46(2), 341–361, doi:10.1111/j.1365–
246X.1976.tb04162.x.
Backus, G., and Mulcahy, M. 1976b. Moment tensors and other phenomenologi-
cal descriptions of seismic sources II. Discontinuous displacements. Geophysical
Journal of the Royal Astronomical Society, 47(2), 301–329, doi:10.1111/j.1365–
246X.1976.tb01275.x.
Baisch, S., Bohnhoff, M., Ceranna, L., Tu, Y., and Harjes, H.-P. 2002. Probing the crust to
9 km depth: fluid injection experiments and induced seismicity at the KTB superdeep
drilling hole, Germany. Bulletin of Seismological Society of America, 92(6), 2369–
2380.
Baisch, S., Voros, R., Weidler, R., and Wyborn, D. 2009. Investigation of fault mechanisms
during geothermal reservoir stimulation experiments in the Cooper Basin, Australia.
Bulletin of the Seismological Society of America, 99(1), 148–158.
Barenblatt, G. I. 1996. Scaling, Self-similarity, and Intermediate Asymptotics. New York:
Cambridge University Press. ISBN-10 0-521-43522-6.
Barenblatt, G. I., Entov, V. M., and Ryzhik, V. M. 1990. Flow of Fluids Through Natural
Rocks. Dordrecht: Kluwer Academic Publishers.
259
260 References
Baria, R., Baumgartner, J., Gerard, A., Jung, R., and Garnish, J. 1999. European HDR
research programme at Soultz-sous-Florets, France 1987–1996. Geothermics, 28, doi:
10.1016/S0375–6505(99)00036-x.
Becker, K., Shapiro, S. A., Stanchits, S., Dresen, G., and Vinciguerra, S. 2007. Stress
induced elastic anisotropy of the Etnean Basalt: theoretical and laboratory exami-
nation. Geophysical Research Letters, 34, L11307.
Beresnev, I. A. 2001. What we can and cannot learn about earthquake sources from the
spectra of seismic waves. Bulletin of the Seismological Society of America, 91, 397–
400, doi:10.1002/jgrb.50362.
Berryman, J. G. 1992. Effective stress for transport properties of inhomogeneous porous
rock. Journal of Geophysical Research, 97, 17409–17424.
Berryman, J. G. 1993. Effective stress rules for pore-fluid transport in rocks contain-
ing two minerals. International Journal of Rock Mechanics and Mining Sciences &
Geomechanics Abstracts, 30, 1165–1168.
Biot, M. A. 1956. Theory of propagation of elastic waves in a fluid-saturated porous
solid. I. Low-frequency range. Journal of the Acoustical Society of America, 28,
168–178.
Biot, M. A. 1962. Mechanics of deformation and acoustic propagation in porous media.
Journal of Applied Physics, 33, 1482–1498.
Bommer, J. J., Oates, S., Cepeda, J. M., et al. 2006. Control of hazard due to seismic-
ity induced by a hot fractured rock geothermal project. Engineering Geology, 83(4),
287–306.
Bouchbinder, E., Goldman, T., and Fineberg, J. 2014. The dynamics of rapid fracture:
instabilities, nonlinearities and length scales. Reports on Progress in Physics, 77(4),
1–30, doi:10.1088/0034–4885/77/4/046501.
Brown, R. J. S., and Korringa, J. 1975. On the dependence of the elastic properties of a
porous rock on the compressibility of the pore fluid. Geophysics, 40, 608–616.
Brune, J. N. 1970. Tectonic stress and the spectra of seismic shear waves from earth-
quakes. Journal of Geophysical Research, 75(26), 4997–5009, doi:10.1029/JB07
5i026p04997.
Buske, S. 1999. 3-D prestack Kirchhoff migration of the ISO89-3D data set. Pure and App-
lied Geophysics, D. Gajewski and W. Rabbel (eds). Basel Burkhäuser, pp. 157–171.
Carcione, J. M. 2007. Wave Fields in Real Media: Wave Propagation in Anisotropic,
Anelastic, Porous and Electromagnetic Media. Oxford: Elsevier.
Carcione, J. M., and Tinivella, U. 2001. The seismic response to overpressure: a mod-
elling study based on laboratory, well and seismic data. Geophysical Prospecting, 49,
523–539.
Carslaw, H. S., and Jaeger, J. C. 1973. Conduction of Heat in Solids. Oxford: Clarendon
Press.
Carter, R. D. 1957. Derivation of the general equation for estimating the extent of the
fractured area. Appendix I of Optimum Fluid Characteristics for Fracture Exten-
sion, G. C. Howard and C. R. Fant (eds.). New York: American Petroleum Institute,
pp. 261–269.
Cerveny, V. 2005. Seismic Ray Theory. Cambridge University Press.
Chandler, R. N., and Johnson, D. L. 1981. The equivalence of quasistatic flow in fluid-
saturated porous media and Biot’s slow wave in the limit of zero frequency. Journal
of Applied Physics, 52(5), 3391–3395.
Cheng, A. H.-D. 1997. Material coefficients of anisotropic poroelasticity. Inter-
national Journal of Rock Mechanics and Mining Sciences, 34(2), 199–205,
doi:10.1016/S0148–9062(96)00055–1.
Ciz, R., and Shapiro, S. A. 2007. Generalization of Gassmann’s equations for porous media
saturated with a solid material. Geophysics, 72, A75–A79.
References 261
Ciz, R., and Shapiro, S. A. 2009. Stress-dependent anisotropy in transversely isotropic

rocks: Comparison between theory and laboratory experiment on shale. Geophysics,
74, D7–D12.
Cleary, M. P. 1977. Fundamental solutions for a fluid-saturated porous solid. International
Journal of Solids and Structures, 13, 785–806.
Cornet, F. 2000. Comment on “Large-scale in situ permeability tensor of rocks from
induced microseismicity” by S. A. Shapiro, P. Audigane and J.-J. Royer. Geophysical
Journal International, 140, 465–469.
Cornet, F. H., Helm, J., Poitrenaud, H., and Etchecopar, A. 1997. Seismic and aseismic slip
induced by large scale fluid-injections. PAGEOPH, 150, 563–583.
Coussy, O. 2004. Poromechanics. Chichester: Wiley.
Crank, J. 1975. The Mathematics of Diffusion. 2nd edn. Oxford: Clarendon Press.
Dahm, T., Becker, D., Bischoff, M., et al. 2013. Recommendation for the discrimina-
tion of human-related and natural seismicity. Journal of Seismology, 17, 197–202,
doi:10.1007/s10950–012–9295–6.
Deichmann, N., and Giardini, D. 2009. Earthquakes induced by the stimulation of an
enhanced geothermal system below Basel (Switzerland). Seismological Research
Letters, 80(5), 784–798, doi:10.1785/gssrl.80.5.784.
Detournay, E., and Cheng, A. H.-D. 1993. Fundamentals of poroelasticity. Chapter 5,
pp. 113–171 of: Hudson, J. A. (ed.), Comprehensive Rock Engineering: Principles,
Practice and Projects. Oxford: Pergamon Press.
Dieterich, J. 1978. Pre-seismic fault slip and earthquake prediction. Journal of Geophysical
Research, 83, 3940–3948.
Dieterich, J. 1994. A constitutive law for rate of earthquake production and its application
to earthquake clustering. Journal of Geophysical Research, 99, 2601–2618.
Dinske, C., and Shapiro, S. A. 2013. Seismotectonic state of reservoirs inferred from mag-
nitude distributions of fluid-induced seismicity. Journal of Seismology, 17, 13–25,
doi:10.1007/s10950–012–9292–9.
Dinske, C., Shapiro, S. A., and Rutledge, J. T. 2010. Interpretation of microseismicity
resulting from gel and water fracturing of tight gas reservoirs. Pure and Applied
Geophysics, 167(1–2), 169–182, doi:10.1007/s00024–009–0003–6.
Duffy, J., and Mindlin, R. D. 1957. Stress–strain relations and vibrations of a granular
medium. Journal of Applied Mechanics, 24, 585–593.
Duncan, P., and Eisner, L. 2010. Reservoir characterization using surface microseismic
monitoring. Geophysics, 75(5), 75A139–75A146, doi:10.1190/1.3467760.
Dyer, B., Jupe, A., Jones, R. H., et al. 1994. Microseismic Results from the European
HDR Geothermal Project at Soultz-sous-Forets, Alsace, France. CSM Associated Ltd,
IR03/24.
Dyer, B. C., Schanz, U., Ladner, F., Häring, M., and Spillman, T. 2008. Microseismic
imaging of a geothermal reservoir stimulation. The Leading Edge, 27(7), 856–869,
doi:10.1190/1.2954024.
Eberhart-Phillips, D., Han, D.-H., and Zoback, M. D. 1989. Empirical relationships
among seismic velocity, effective pressure, porosity and clay content in sandstone.
Geophysics, 54, 82–89.
Economides, M. J., and Nolte, K. G. (eds.) 2003. Reservoir Stimulation, 3rd edn.
Chichester: Wiley, pp. 5-1–5-14.
Edelman, I. Y., and Shapiro, S. A. 2004. An analytical approach to the description of
fluid-injection induced microseismicity in porous rock. Doklady Earth Sciences, 399,
1108–1112.
Emmermann, R., and Lauterjung, J. 1997. The German Continental Deep Drilling Program
KTB: overview and major results. Journal of Geophysical Research: Solid Earth,
102(B8), 18 179–18 201, doi:10.1029/96JB03945.
262 References
Erzinger, J., and Stober, I. 2005. Introduction to special issue: long-term fluid production
in the KTB pilot hole. Geofluids, 5, 1–7.
Fehler, M., House, L., Phillips, W. S., and Potter, R. 1998. A method to allow tempo-
ral variation of velocity in travel-time tomography using microearthquakes induced
during hydraulic fracturing. Tectonophysics, 289, 189–202.
Ferreira, J. M., Oliveira, R. T. De, Assumpcao, M., et al. 1995. Correlation of seismicity
and water level in the Acu Reservoir – an example from northeast Brazil. Bulletin of
the Seismological Society of America, 85(5), 1483–1489.
Fischer, T., Hainzl, S., Eisner, L., Shapiro, S. A., and Le Calvez, J. 2008. Microseismic sig-
natures of hydraulic fracture growth in sediment formations: observations and mod-
eling. Journal of Geophysical Research, 113(B02307), doi:10.1029/2007JB005070.
Fisher, M. K., Davidson, B. M., Goodwin, A. K., et al. 2002. Integrating fracture mapping
technologies to optimize stimulations in the Barnett Shale. Paper SPE 77411.
Fisher, M. K., Heinze, J. R., Harris, C. D, et al. 2004. Optimizing horizontal comple-
tion techniques in the Barnett Shale using microseismic fracture mapping. Paper SPE
90051.
Fletcher, J. B., and Sykes, L. R. 1977. Earthquakes related to hydraulic mining and natural
seismic activity in Western New York State. Journal of Geophysical Research, 82,
3767–3780.
Frenkel, J. 2005. On the theory of seismic and seismoelectric phenomena in a moist soil.
Journal of Engineering Mechanics, 131, 879–887.
Freund, D. 1992. Ultrasonic compressional and shear velocities in dry clastic rocks as
a function of porosity, clay content, and confining pressure. Geophysical Journal
International, 108, 125–135.
Gangi, A. F. 1978. Variation of whole and fractured porous rock permeability with con-
fining stress. International Journal of Rock Mechanics and Mining Science, 15,
249–257.
Gangi, A. F., and Carlson, R. L. 1996. An asperity-deformation model for effective
pressure. Tectonophysics, 256, 241–251.
Gassmann, F. 1951. Über die Elastizität poröser Medien. Vierteljahresschrift der Natur-
forschenden Gesellschaft in Zurich, 96, 1–23.
Gavrilenko, P., Singh, Chandrani, and Chadha, R. K. 2010. Modelling the hydromechan-
ical response in the vicinity of the Koyna reservoir (India): results for the initial
filling period. Geophysical Journal International, 183, 461–477, doi:10.1111/j.1365–
246X.2010.04752.x.
Gelinsky, S., and Shapiro, S. A. 1997. Dynamic-equivalent medium approach for thinly
layered saturated sediments. Geophysical Journal International, 128, F1–F4.
Giardini, D. 2009. Geothermal quake risks must be faced. Nature, 462(7275), 848–849,
doi:10.1038/462848a.
Giardini, D., Grünthal, G., Shedlock, K. M., and Zhang, P. 2003. The GSHAP Global Seis-
mic Hazard Map. International Handbook of Earthquake & Engineering Seismology.
International Geophysics Series, vol. 81, no. B. Amsterdam: Academic Press, pp.
1233–1239.
Goertz-Allmann, B. P., Goertz, A., and S., Wiemer. 2011. Stress drop variations of induced
earthquakes at the Basel geothermal site. Geophysical Research Letters, 38, L09308,
doi:10.1029/2011GL047498.
Goulty, N. R. 1998. Relationship between porosity and effective stress in shales. First
Break, 16, 413–419.
Gräsle, W., Kessels, W., Kümpel, H.-J., and Li, X. 2006. Hydraulic observations from a
1 year fluid production test in the 4000 m deep KTB pilot borehole. Geofluids, 6,
doi:10.1111/j.1468–8123.2006.00124.x.
References 263
Grechka, V. 2009. Applications of Seismic Anisotropy in the Oil and Gas Industry. Houten:
EAGE Publications.
Grechka, V., and Yaskevich, S. 2013. Inversion of microseismic data for triclinic velocity
models. Geophysical Prospecting, 61(6), 1159–1170, doi:10.1111/1365–2478.12042.
Griffith, A. A. 1921. The phenomena of rupture and flow in solids. Philosophical
Transactions of the Royal Society of London, A221, 163–198.
Griffith, A. A. 1924. The theory of rupture. Proceedings of 1st International Congress on
Applied Mechanics, Delft, pp. 55–63.
Grünthal, G., and Wahlström, R. 2003. An Mw based earthquake catalogue for cen-
tral, northern and northwestern Europe using a hierarchy of magnitude conversions.
Journal of Seismology, 7, 507–531.
Guha, S. K. 2000. Induced Earthquakes. London: Kluwer Academic, Dordrecht.
Gupta, H. K. 2002. A review of recent studies of triggered earthquakes by artificial
water reservoirs with special emphasis on earthquakes in Koyna, India. Earth-Science
Reviews, 58, 279–310.
Gurevich, B. 2004. A simple derivation of the effective stress coefficient for seismic
velocities in porous rocks. Geophysics, 69, 393–397, doi:10.1190/1.1707058.
Gurevich, B., and Lopatnikov, S. 1995. Velocity and attenuation of elastic waves in finely
layered porous rocks. Geophysical Journal International, 121, 933–947.
Gurevich, B., Makarynska, D., de P. O. Bastos, and Pervukhina, M. 2010. A simple
model for squirt-flow dispersion and attenuation in fluid-saturated granular rocks.
Geophysics, 75(6), N109–N120, doi:10.1190/1.3509782.
Gutenberg, B., and Richter, C. F. 1954. Seismicity of Earth and Associated Phenomenon.
Princeton: Princeton University Press.
Hainzl, S., Fischer, T., and Dahm, T. 2012. Seismicity-based estimation of the driving
fluid pressure in the case of swarm activity in Western Bohemia. Geophysical Journal
International, 191(1), 271–281, doi:10.1111/j.1365–246X.2012.05610.x.
Häring, M., Schanz, U., Ladner, F., and Dyer, B. 2008. Characterisation of the Basel
1 enhanced geothermal system. Geothermics, 37(5), 469–495, doi:10.1016/j.geoth
ermics.2008.06.002.
Harjes, H.-P., Bram, K., Dürbaum, H.-J., et al. 1997. Origin and nature of crustal reflec-
tions: Results from integrated seismic measurements at the KTB superdeep drilling
site. Journal of Geophysical Research, 102(B8), 18 267–18 288.
Hashin, Z., and Strikman, S. 1963. A variational approach to the elastic behavior of
multiphase materials. Journal of the Mechanics and Physics of Solids, 11, 127–140.
Hirschmann, G., and Lapp, M. 1995. Evaluation of the structural geology of the KTB
Hauptbohrung (KTB-Oberpfalz HB). KTB Report 94-1. Niedersächsisches Lan-
desamt für Bodenforschung, pp. 285–308.
House, L. 1987. Locating microearthquakes induced by hydraulic fracturing in crystalline
rocks. Geophysical Research Letters, 14, 919–921.
Hudson, D. J. 1964. Lectures on Elementary Statistics and Probability. Geneva: CERN,
European Organization for Nuclear Research.
Hudson, J. A., Pointer, T., and Liu, E. 2001. Effective-medium theories for fluid-saturated
materials with aligned cracks. Geophysical Prospecting, 49, 509–522.
Hummel, N., and Müller, T. M. 2009. Microseismic signatures of non-linear pore-
fluid pressure diffusion. Geophysical Journal International, 179(3), 1558–1565,
doi:10.1111/j.1365–246X.2009.04373.x.
Hummel, N., and Shapiro, S. A. 2012. Microseismic estimates of hydraulic diffusivity in
case of non-linear fluid-rock interaction. Geophysical Journal International, 188(3),
1441–1453, doi:10.1111/j.1365–246X.2011.05346.x.
264 References
Hummel, N., and Shapiro, S. 2013. Nonlinear diffusion-based interpretation of induced

microseismicity: a Barnett Shale hydraulic fracturing case study. Geophysics, 78(5),
B211–B226, doi:10.1190/geo2012–0242.1.
Jaeger, J. C., Cook, N. G. W., and Zimmerman, R. W. 2007. Fundamentals of Rock
Mechanics. Blackwell Publishing.
Jones, S. M. 1995. Velocities and quality factors of sedimentary rocks at low and high
effective pressures. Geophysical Journal International, 123, 774–780.
Jost, M. L., Büßelberg, T., Jost, Ö., and Harjes, H.-P. 1998. Source parameters of injection-
induced microearthquakes at 9 km depth at the KTB Deep Drilling site, Germany.
Bulletin of the Seismological Society of America, 88(3), 815–832.
Jung, R., Cornet, F., Rummel, F., and Willis-Richard, J. 1996. Hydraulic stimulation results
1992/1993. Pages 31–41 of: Baria, R., Baumgärtner, J., and Gérard, A. (eds.), Euro-
pean Hot Dry Rock Programme, 1992–1995. Extended Summary of the Final Report
to European Community (DG XII), Contract N JOU2-CT92-0115 (Date of issue: 28
March 1996, Version 1).
Kaieda, H., and Sasaki, S. 1998. Development of fracture evaluation methods for Hot Dry
Rock geothermal power – Ogachi reservoir evaluation by the AE method. CRIEPI
report U97107 (in Japanese with English abstract).
Kaieda, H., Kiho, K., and Motojima, I. 1993. Multiple fracture creation for hot dry rock
development. Trends in Geophysical Research, 2, 127–139.
Kalpna, and Chander, R. 2000. Greens function based stress diffusion solutions in the
porous elastic half space for time varying finite reservoir. Physics of the Earth and
Planetary Interiors, 120, 93–101, doi:10.1002/jgrb.50362.
Kanamori, H. 1977. The energy release in great earthquakes. Journal of Geophysical
Research, 82(20), 2981–2987, doi:10.1029/JB082i020p02981.
Kanamori, H., and Brodsky, E. E. 2004. The physics of earthquakes. Reports on Progress
in Physics, 67, 1429–1496.
Karpfinger, F., Müller, T. M., and Gurevich, B. 2009. Green’s functions and radiation
patterns in poroelastic solids revisited. Geophysical Journal International, 178(1),
327–337, doi:10.1111/j.1365–246X.2009.04116.x.
Kaselow, A., and Shapiro, S. A. 2004. Stress sensitivity of elastic moduli and electrical
resistivity in porous rocks. Journal of Geopysics and Engineering, 1, 1–11.
Kaselow, A., Becker, K., and Shapiro, S. A. 2006. Stress sensitivity of seismic and electric
rock properties of the upper continental crust at the KTB. PAGEOPH, 163, 1021–
1029, doi:10.1007/s00024–006–0063–9.
Kern, H., and Schmidt, R. 1990. Physical properties of the KTB core samples at simulated
in situ conditions. Scientific Drilling, 1, 217–223.
Kern, H., Popp, T., and Schmidt, R. 1994. The effect of deviatoric stress on the rock prop-
erties: an experimental study simulating the in-situ stress field at the KTB drilling site,
Germany. Surveys in Geophysics, 15, 467–479.
Khaksar, A., Griffiths, C. M., and McCann, C. 1999. Compressional- and shear-wave
velocities as a function of confining stress in dry sandstones. Geophysical Prospect-
ing, 47, 487–508.
Kirstetter, O., and MacBeth, C. 2001. Compliance-based interpretation of dry frame
pressure sensitivity in shallow marine sandstone. Expanded Abstracts. Society of
Exploration Geophysicists, San Antonio, pp. 2132–2135.
Klee, G., and Rummel, F. 1993. Hydrofrac stress data for the european HDR research
project test site Soultz-sous-Forêts. International Journal of Rock Mechanics and
Mining Sciences & Geomechanics Abstracts, 30(7), 973–976.
References 265
Kummerow, J. 2010. Using the value of the crosscorrelation coefficient to locate micro-
seismic events. Geophysics, 75(4), MA47–MA52.
Kummerow, J. 2013. Joint arrival time optimization for microseismic events recorded
by seismic borehole arrays. doi:10.3997/2214–4609.20130401 Extended Abstracts of
75th EAGE Conference & Exhibition, Session: Microseismic Event Characterisation.
Kummerow, J., Reshetnikow, A., Häring, M., and Asanuma, H. 2012. Distribution of the
Vp/Vs ratio within the Basel 1 Geothermal Reservoir from microseismic data. In:
Extended Abstracts of 74th EAGE Conference & Exhibition, Session: Microseismic
Interpretation.
Kümpel., H.-J, Erzinger, J., and Shapiro, S. A. 2006. Two massive hydraulic tests
completed in deep KTB pilot hole. Scientific Drilling, (3) pp. 40–42.
Landau, L. D., and Lifshitz, E. M. 1987. Theory of elasticity (in Russian). Moscow: Nauka,
Glavnaja Redaktsija Phys.-Math. Lit.
Landau, L. D., and Lifshitz, E. M. 1991. Lehrbuch der theoretischen Physik (in German),
5th edn. Vol. VI. Moscow: Acadamie, Berlin, Germany.
Langenbruch, C., and Shapiro, S. A. 2010. Decay rate of fluid-induced seismicity after
termination of reservoir stimulations. Geophysics, 75, doi:10.1190/1.3506005.
Langenbruch, C., and Shapiro, S. A. 2014. Gutenberg–Richter relation originates from
Coulomb stress fluctuations caused by elastic rock heterogeneity. Journal of Geo-
physical Research: Solid Earth, 119(2), 1220–1234, doi:10.1002/2013JB010282.
Langenbruch, C., Dinske, C., and Shapiro, S. A. 2011. Inter event times of fluid induced
earthquakes suggest their Poisson nature. Geophysical Research Letters, 38(21),
doi:10.1029/2011GL049474.
Lay, T., and Wallace, T. C. 1995. Modern Global Seismology. Academic Press.
Li, M., Bernabé, Y., Xiao, W.-I., Chen, Z.-Y., and Liu, Z.-Q. 2009. Effective pressure law
for permeability of E-bei sandstones. Journal of Geophysical Research: Solid Earth,
114(B7), doi:10.1029/2009JB006373.
Lin, G., and Shearer, P. 2007. Estimating local V p /Vs ratios within similar earthquake
clusters. Bulletin of the Seismological Society of America, 97(2), 379–388.
Lomax, A., Michelini, A., and Curtis, A. 2009. Earthquake location, direct, global-search
methods. Pages 2449–2473, doi:10.1007/978–0–387–30440–3–150 Meyers, R. A.
(ed.), Encyclopedia of Complexity and Systems Science. New York: Springer.
Lopatnikov, S. L., and Cheng, A. H.-D. 2005. If you ask a physicist from any country: a
tribute to Yacov Il’ich Frenkel. Journal of Engineering Mechanics, 131, 875–878.
Madariaga, R. 1976. Dynamics of an expanding circular fault. Bulletin of the Seismological
Society of America, 66(3), 639–666.
Majer, E. L., Baria, R., Stark, M., et al. 2007. Induced seismicity associ-
ated with Enhanced Geothermal Systems. Geothermics, 36, 185–222,
doi:10.1016/j.geothermics.2007.03.003.
Mandelis, A. 2000. Diffusion waves and their uses. Physics Today, 53(8), 29–34.
Mavko, G., and Jizba, D. 1991. Estimating grain-scale fluid effects on velocity dispersion
in rocks. Geophysics, 12, 1940–1949.
Mavko, G., Mukerji, T., and Dvorkin, J. 1998. The Rock Physics Handbook: Tools for
Seismic Analysis in Porous Media. Cambridge University Press.
Maxwell, S. 2014. Microseismic Imaging of Hydraulic Fracturing: Improved Engineering
of Unconventional Shale Reservoirs. SEG, Tulsa.
Maxwell, S. C., Waltman, C., Warpinski, N. R., Mayerhofer, M. J., and Boroumand,
N. 2009. Integrating fracture-mapping technologies to improve stimulations in
the Barnett Shale. SPE Reservoir Evaluation & Engineering, 12(1), 48–52,
doi:10.2118/102801–PA.
266 References
Maxwell, S. C., Waltman, C. K., Warpinski, N. R., Mayerhofer, M. J., and Boroumand, N.
2006. Imaging seismic deformation induced by hydraulic fracture complexity. Paper
SPE 102801.
McClintock, F. A., and Walsh, J. B. 1962. Friction on Griffith cracks in rocks under pres-
sure. Pages 1015–1021 of: Proceedings of 4th US National Congress on Applied
Mechanics.
McGarr, A. 1976. Seismic moments and volume changes. Journal of Geophysical
Research, 81(8), 1487–1494, doi:10.1029/JB081i008p01487.
McGarr, A., Simpson, D., and Seeber, L. 2002. Case histories of induced and triggered
seismicity. Chapter 40, pages 647–665 of: Lee, W. H. K. et al. (eds.), International
Handbook of Earthquake and Engineering Seismology, International Geophysics
Series, vol. 81A. New York: Elsevier.
Merkel, R. H., Barree, R. D., and Towle, G. 2001. Seismic response of Gulf of Mexico
reservoir rocks with variations in pressure and water saturation. The Leading Edge,
20, 290–299.
Millich, E., Neugebauer, H. J., Huenges, E., and Nover, G. 1998. Pressure dependence of
permeability and Earth tide induced fluid flow. Geophysical Research Letters, 25(6),
809–812, doi:10.1029/98GL00395.
Mindlin, R. D. 1949. Compliance of elastic bodies in contact. Journal of Applied
Mechanics, 16, 259–268.
Mukuhira, Y., Asanuma, H., Niitsuma, H., Schanz, U., and Häring, M. 2008. Characteri-
zation of microseismic events with larger magnitude collected at Basel, Switzerland
in 2006. GRC Transactions, 32, 87–93.
National Research Council. 2013. Induced Seismicity Potential in Energy Technologies.
Washington, DC: The National Academies Press.
Norris, A. 1989. Stoneley-wave attenuation and dispersion in permeable formations.
Geophysics, 54, 330–341.
Norris, A. 1993. Low-frequency dispersion and attenuation in partially saturated rocks.
Journal of the Acoustical Society of America, 94, 359–370.
Nur, A., and Booker, J. 1972. Aftershocks caused by pore fluid flow? Science, 175,
885–887.
Oelke, A., Alexandrov, D., Abakumov, I., et al. 2013. Seismic reflectivity of
hydraulic fractures approximated by thin fluid layers. Geophysics, 78(4), T79–T87,
doi:10.1190/geo2012–0269.1.
Ohtake, M. 1974. Seismic activity induced by water injection at Matsushiro, Japan. Journal
of Physics of the Earth, 22, 163–176.
Parotidis, M., Rothert, E., and Shapiro, S. A. 2003. Pore-pressure diffusion: A possible trig-
gering mechanism for the earthquake swarms 2000 in Vogtland/NW-Bohemia, central
Europe. Geophysical Research Letters, 30(20), 2075, doi:10.1029/2003GL018110.
Parotidis, M., Shapiro, S. A., and Rothert, E. 2004. Back front of seismicity induced
after termination of borehole fluid injection. Geophysical Research Letters, 31,
doi:10.1029/2003GL018987.
Parotidis, M., Shapiro, S. A., and Rothert, E. 2005. Evidence for triggering of the Vogt-
land swarms 2000 by pore pressure diffusion. Journal of Geophysical Research,
110(B05S10), 1–12, doi:10.1029/2004JB003267.
Pavlis, L. 1986. Appraising earthquake hypocenter location errors: a complete practical
approach for single-event locations. Bulletin of the Seismological Society of America,
76, 1699–1717.
Pearson, C. 1981. The relationship between microseismicity and high pore pressures dur-
ing hydraulic stimulation experiments in low permeability granitic rocks. Journal of
Geophysical Research, 86, 7855–7864.
References 267
Peirce, A., and Detournay, E. 2008. An implicit level set method for modeling hydrauli-
cally driven fractures. Computer Methods in Applied Mechanics and Engineering,
197(3340), 2858–2885, doi:10.1016/j.cma.2008.01.013.
Pervukhina, M., Gurevich, B., Dewhurst, D. N., and Siggins, A. F. 2010. Applicability of
velocity–stress relationships based on the dual porosity concept to isotropic porous
rocks. Geophysical Journal International, 181, 1473–1479, doi:10.1111/j.1365–
246X.2010.04535.x.
Phillips, W. S., House, L. S., and Fehler, M. C. 1997. Detailed joint structure in a
geothermal reservoir from studies of induced microearthquake clusters. Journal of
Geophysical Research, 102(B6), 11 745–11 763, doi:10.1029/97JB00762.
Prasad, M., and Manghnani, M. H. 1997. Effects of pore and differential pressure on
compressional wave velocity and quality factor in Berea and Michigan sandstones.
Geophysics, 62, 1163–1176.
Pride, S. R., Berryman, J. G., and Harris, J. M. 2004. Seismic attenuation due
to wave-induced flow. Journal of Geophysical Research: Solid Earth, 109(B1),
doi:10.1029/2003JB002639.
Rentsch, S., Buske, S., Lüth, S., and Shapiro, S. A. 2007. Fast location of seismicity: a
migration-type approach with application to hydraulic-fracturing data. Geophysics,
72, S33–S40, doi:10.1190/1.2401139.
Rentsch, S., Buske, S., Gutjahr, S., Kummerow, J., and Shapiro, S. A. 2010. Migration-
based location of seismicity recorded with an array installed in the main hole
of the San Andreas Fault Observatory at Depth (SAFOD). Geophysical Journal
International, 182(1), 477–492, doi:10.1111/j.1365–246X.2010.04638.x.
Reshetnikov, A., Buske, S., and Shapiro, S. A. 2010. Seismic imaging using microseismic
events: results from the San Andreas Fault System at SAFOD. Journal of Geophysical
Research: Solid Earth, 115(B12), doi:10.1029/2009JB007049.
Reshetnikov, A., Kummerow, J., Shapiro, S. A., Asanuma, H., and Häring, M.
2013. Microseismic reflection imaging of stimulated reservoirs and fracture zones.
doi:10.3997/2214–4609.20131279 Extended Abstracts of 75th EAGE Conference &
Exhibition – Workshops Session: WS16 – Micro Seismicity – What Now? What Next?.
Rice, J. R. 1980. The mechanics of earthquake rupture, course 78, 1979, pages 556–649
of: Dziewonski, A. M., and Boschi, E. (eds.), Physics of the Earth’s Interior (Proc.
International School of Physics “Enrico Fermi”). Italian Physical Society and North-
Holland Publ. Co.
Rice, J. R., and Cleary, M. P. 1976. Some basic stress diffusion solutions for fluid-saturated
elastic porous media with compressible constituents. Reviews of Geophysics and
Space Physics, 14, 227–241.
Richards, P. G., Waldhauser, F., Schaff, D., and Kim, W.-Y. 2006. The applicability of
modern methods of earthquake location. Pure and Applied Geophysics, 163(2–3),
351–372, doi:10.1007/s00024–005–0019–5.
Roeloffs, E. A. 1988. Fault stability changes induced beneath a reservoir with cyclic
variations in water level. Journal of Geophysical Research: Solid Earth, 93(B3),
2107–2124, doi:10.1029/JB093iB03p02107.
Rothert, E., and Shapiro, S. A. 2003. Microseismic monitoring of borehole fluid injections:
data modeling and inversion for hydraulic properties of rocks. Geophysics, 68(2),
685–689, doi:10.1190/1.1567239.
Rothert, E., and Shapiro, S. A. 2007. Statistics of fracture strength and fluid-
induced microseismicity. Journal of Geophysical Research, 112, B04309,
doi:10.1029/2005JB003959.
268 References
Rudnicki, J. W. 1986. Fluid mass sources and point forces in linear elastic diffusive solids.
Mechanics of Materials, 5, 383–393.
Ruina, A. 1983. Slip instability and state variable friction laws. Journal of Geophysical
Research, 88, 359–370.
Rutledge, J. T., and Phillips, W. S. 2003. Hydraulic stimulation of natural fractures as
revealed by induced microearthquakes, Carthage Cotton Valley gas field, east Texas.
Geophysics, 68(2), 441–452, doi:10.1190/1.1567214.
Rutledge, J. T., Phillips, W. S., and Mayerhofer, M. J. 2004. Faulting induced by forced
fluid injection and fluid flow forced by faulting: An interpretation of hydraulic-
fracture microseismicity, Carthage Cotton Valley gas field, Texas. Bulletin of the
Seismological Society of America, 94, 1817–1830.
Rytov, S. M., Kravtsov, Yu. A., and Tatarskii, V. I. 1989. Wave Propagation Through
Random Media. Principles of statistical radiophysics, vol. 4. Berlin: Springer Verlag.
Schoenball, M., Müller, T. M., Müller, B. I. R., and Heidbach, O. 2010. Fluid-induced
microseismicity in pre-stressed rock masses. Geophysical Journal International,
180(2), doi:10.1111/j.1365–246X.2009.04443.x.
Scholz, C. H. 2002. The Mechanics of Earthquakes and Faulting, 2nd edn. Cambridge:
Cambridge University Press.
Scholze, M., Stürmer, K., Kummerow, J., and Shapiro, S. A. 2010. An approach to analyse
microseismic event similarity. In: Extended Abstracts of 72nd EAGE Conference &
Exhibition, Session: Passive & Microseismic Methods (EAGE).
Segall, P. 1989. Earthquakes triggered by fluid extraction. Geology, 17(10), 942–946,
doi:10.1130/0091–7613(1989).
Segall, P. 2010. Earthquake and Volcano Deformation. Princeton: Princeton University
Press.
Segall, P., and Fitzgerald, S. 1998. A note on induced stress changes in hydrocarbon and
geothermal reservoirs. Tectonophysics, 289, 117–128.
Segall, P., and Rice, J. R. 1995. Dilatancy, compaction, and slip instability of a fluid-
infiltrated fault. Journal of Geophysical Research, 100, 22 155–22 172.
Segall, P., Desmarais, E., Shelly, D., Miklius, A., and Cervelli, P. 2006. Earthquakes
triggered by silent slip events on Kilauea volcano, Hawaii. Nature, 442, 71–74.
Shapiro, S. A. 1992. Elastic waves scattering and radiation by fractal inhomogeneity of a
medium. Geophysical Journal International, 110, 591–600.
Shapiro, S. A. 2000. An inversion for fluid transport properties of three-dimensionally het-
erogeneous rocks using induced microseismicity. Geophysical Journal International,
143, 931–936.
Shapiro, S. A. 2003. Elastic piezosensitivity of porous and fractured rocks. Geophysics,
68, 482–486, doi: 10.1190/1.1567215.
Shapiro, S. A. 2008. Microseismicity: a Tool for Reservoir Characterization. Houten, the
Netherlands: EAGE Publications.
Shapiro, S. A., and Dinske, C. 2009a. Fluid-induced seismicity: pressure diffusion and
hydraulic fracturing. Geophysical Prospecting, 57, 301–310, doi:10.1111/j.1365–
2478.2008.00770.xi.
Shapiro, S. A., and Dinske, C. 2009b. Scaling of seismicity induced by nonlin-
ear fluid-rock interaction. Journal of Geophysical Research, 114(B9), B09307,
doi:10.1029/2008JB006145.
Shapiro, S. A., and Fayzullin, I. S. 1992. Fractal properties of fault systems by scattering
of body seismic waves. Tectonophysics, 202, 177–181.
Shapiro, S. A., and Hubral, P. 1999. Elastic Waves in Random Media. Fundamentals of
Seismic Stratigraphic Filtering. Berlin: Springer Verlag.
References 269
Shapiro, S. A., and Kaselow, A. 2005. Porosity and elastic anisotropy of rocks
under tectonic stress and pore-pressure changes. Geophysics, 70, N27–N38,
doi:10.1190/1.2073884.
Shapiro, S. A., Huenges, E., and Borm, G. 1997. Estimating the permeability from
fluid-injection-induced seismic emissions at the KTB site. Geophysical Journal
International, 131, F15–F18.
Shapiro, S. A., Audigane, P., and Royer, J.-J. 1999. Large-scale in situ permeability ten-
sor of rocks from induced microseismicity. Geophysical Journal International, 137,
207–213.
Shapiro, S. A., Rothert, E., Rath, V., and Rindschwentner, J. 2002. Characterization of
fluid transport properties of reservoirs using induced microseismicity. Geophysics,
67, 212–220, doi:10.1190/1.1451597.
Shapiro, S. A., Patzig, R., Rothert, E., and Rindschwentner, J. 2003. Triggering of micro-
seismicity due to pore-pressure perturbation: permeability related signatures of the
phenomenon. PAGEOPH, 160, 1051–1066.
Shapiro, S. A., Rentsch, S., and Rothert, E. 2005a. Characterization of hydraulic prop-
erties of rocks using probability of fluid-induced microearthquakes. Geophysics, 70,
F27–F34, doi:10.1190/1.1897030.
Shapiro, S. A., Rentsch, S., and Rothert, E. 2005b. Fluid-induced seismicity: theory,
modeling, and applications. Journal of Engineering Mechanics, 131, 947–952.
Shapiro, S. A., Kummerow, J., Dinske, C., et al. 2006a. Fluid induced seismicity guided by
a continental fault: injection experiment of 2004/2005 at the German Deep Drilling
Site (KTB). Geophysical Research Letters, 33(1), L01309, doi:10.1029/2005GL024
659.
Shapiro, S. A., Dinske, C., and Rothert, E. 2006b. Hydraulic-fracturing controlled
dynamics of microseismic clouds. Geophysical Research Letters, 33, L14312,
doi:10.1029/2006GL026365.
Shapiro, S. A., Dinske, C., and Kummerow, J. 2007. Probability of a given-magnitude
earthquake induced by a fluid injection. Geophysical Research Letters, 34, L22314,
doi:10.1029/2007GL031615.
Shapiro, S. A., Dinske, C., Langenbruch, C., and Wenzel, F. 2010. Seismogenic index and
magnitude probability of earthquakes induced during reservoir fluid stimulations. The
Leading Edge, 29(3), 304–309, doi:10.1190/1.3353727.
Shapiro, S.A., Krüger, O. S., Dinske, C., and Langenbruch, C. 2011. Magnitudes of
induced earthquakes and geometric scales of fluid-stimulated rock volumes. Geo-
physics, 76, WC53–WC61, doi:10.1190/GEO2010–0349.1.
Shapiro, S. A., Krüger, O. S., and Dinske, C. 2013. Probability of inducing given-
magnitude earthquakes by perturbing finite volumes of rocks. Journal of Geophysical
Research: Solid Earth, 118(7), 3557–3575, doi:10.1002/jgrb.50264.
Shearer, P. M. 1994. Global seismic event detection using a matched filter on long-period
seismograms. Journal of Geophysical Research: Solid Earth, 99(B7), 13 713–13 725,
doi:10.1029/94JB00498.
Shearer, P. M. 2009. Introduction to Seismology. Cambridge: Cambridge University
Press.
Shelly, D. R., Hill, D. P., Massin, F., et al. 2013. A fluid-driven earthquake swarm on
the margin of the Yellowstone caldera. Journal of Geophysical Research, 188, 4872–
4886, doi:10.1002/jgrb.50362.
Simon, M., Gebrande, H., and Bopp, M. 1996. Pre-stack migration and true-amplitude pro-
cessing of DEKORP near-normal incidence and wideangle reflection measurements.
Tectonophysics, 264, 381–392.
270 References
Soma, N., Asanuma, H., Kaieda, H., et al. 2004. On site mapping of microseis-
micity at Cooper Basin, Australia HDR project by the Japanese team. In: Pro-
ceedings, Twenty-ninth Workshop on Geothermal Reservoir Engineering, Stanford
University.
Stober, I., and Bucher, K. 2005. The upper continental crust, an aquifer and its fluid:
hydraulic and chemical data from 4 km depth in fractured crystalline basement rocks
at the KTB test site. Geofluids, 5(1), 8–19, doi:10.1111/j.1468–8123.2004.00106.x.
Talwani, P. 1997. On the nature of reservoir-induced seismicity. PAGEOPH, 150, 473–492.
Talwani, P., and Acree, S. 1985. Pore pressure diffusion and the mechanism of reservoir-
induced seismicity. PAGEOPH, 122, 947–965.
Terzaghi, K. 1936. The shear resistance of saturated soils. Proceedings for the 1st Interna-
tional Conference on Soil Mechanics and Foundation Engineering (Cambridge, MA),
vol. 1, pp. 54–56.
Thomsen, L. 1986. Weak elastic anisotropy. Geophysics, 51(10), 1954–1966.
Thomsen, L. 1995. Elastic anisotropy due to aligned cracks in porous rocks. Geophysical
Prospecting, 43, 805–830.
Thurber, H. C., and Rabinowitz, N. 2000. Advances in Seismic Event Location. Kluwer
Academic Publishers.
Tsvankin, I. 2005. Seismic Signatures and Analysis of Reflection Data in Anisotropic
Media, Volume 29 (Handbook of Geophysical Exploration: Seismic Exploration).
Oxford: Elsevier Science.
Turcotte, D. L., Holliday, J. R., and Rundle, J. B. 2007. BASS, an alternative to ETAS.
Geophysical Research Letters, 34, L12303, doi:10.1029/2007GL029696.
Urbancic, T., and Baig, A. 2013. Validating engineering objectives of hydraulic fracture
stimulations using microseismicity. First Break, 31(7), 73–90.
Van der Kamp, G., and Gale, J. E. 1983. Theory of Earth tide and barometric effects
in porous formations with compressible grains. Water Resources Research, 19,
538–544.
Waldhauser, F., and Ellsworth, W.L. 2000. A double-difference earthquake location algo-
rithm: method and application to the northern Hayward fault, California. Bulletin of
the Seismological Society of America, 90, 1353–1368.
Walsh, J. B. 1981. Effect of pore pressure and confining pressure on fracture permeabil-
ity. International Journal of Rock Mechanics and Mining Sciences & Geomechanics
Abstracts, 18(5), 429–435.
Wang, H. F. 2000. Theory of Linear Poroelasticity with Applications to Geomechanics and
Hydrogeology. Princeton: Princeton University Press.
Warpinski, N., Wolhart, S., and Wright, C. 2001. Analysis and prediction of microseismic-
ity induced by hydraulic fracturing. Paper SPE 71649.
White, J. E. 1983. Underground Sound: Application of Seismic Waves. Amsterdam:
Elsevier.
Yu, C. P., Reshetnikov, A., and Shapiro, S.A. 2013. Simultaneous inversion of anisotropic
velocity model and microseismic event location – synthetic and real data examples.
doi:10.3997/2214–4609.20130407 Extended Abstracts of 75th EAGE Conference &
Exhibition, Session: Microseismic Event Characterisation.
Zarembo, L. K., and Krasilnikov, V. A. 1966. Introduction into Non-linear Acoustics (in
Russian). Moscow: Nauka, Glavnaja Redaktsija Phys.-Math. Lit.
Zimmerman, R. W., Somerton, W. H., and King, M. S. 1986. Compressibility of porous
rocks. Journal of Geophysical Research, 91, 12 765–12 777.
Zoback, M. D. 2010. Reservoir Geomechanics. Cambridge: Cambridge University
Press.
References 271
Zoback, M. D., and Gorelick, S. M. 2012. Earthquake triggering and large-scale geologic
storage of carbon dioxide. Proceedings of the National Academy of Sciences USA,
109, 10 164–10 168, doi: 10.1073/pnas.1202473109.
Zoback, M. D., and Harjes, H.-P. 1997. Injection-induced earthquakes and crustal stress at
9 km depth at the KTB deep drilling site, Germany. Journal of Geophysical Research,
102(B8), 18 477–18 491.
Index
a- and b-values of Gutenberg–Richter law, 205, 240, bulk modulus

243 of grain material, 58
adiabatic elastic constants, 10 of skeleton, 58, 61
aftershocks of earthquake, 206, 235, 236 of undrained material, 59, 61
aging law, 21 of rock, 9, 61, 105, 106, 108
Amontons’ law, 17
angular frequency, 11 characteristic frequency of global flow, 68, 69
anisotropy characteristic frequency of Poiseuille flow, 64, 69
anisotropic elastic media, 8, 10, 11, 43, 65 characteristic magnitude, 239
anisotropic poroelastic media, 140 Christoffel equation, 11
anisotropy parameters, elastic, 43 coefficient of internal friction, 17, 20
anisotropy of hydraulic diffusivity, hydraulic cohesion, 18
anisotropy, 124, 134, 153, 154, 162 compatibility equations, 2
apparent Gutenberg–Richter distribution, 245 completeness magnitude, 208, 209
apparent hydraulic diffusivity, 172 compliance tensor, 5, 6, 51–53, 57, 103
aspect ratio, 104 compliance tensor of undrained rock, 60
attenuation coefficient, 71, 139 compliant porosity, 103–105, 111, 115
average confining normal stress, 58 complimentary error function, 85
compressional tectonic regime, 100
b-value, 215, 222, 240, 243 compressive stress, 4, 100
back front of induced seismicity, 148, 152, 162, 168, confining pressure, 51, 61
172, 175 confining stress, 49–60, 102, 113
back front of seismicity, x, 147, 177 connected porosity, 49, 114–116
Barnett Shale case study, 177, 178, 208 conservation of the fluid mass, 63, 183, 203
Basel case study, 208, 221, 225 continuity equation, 63, 76, 143, 183
biharmonic equation, 14 contracted notation, 6, 60
Biot modulus, 58 Cooper Basin case study, 208, 211
Biot–Willis coefficient, 58, 110, 184 corner frequency, 35
borehole fluid injections, x Cotton Valley case study, 169
bottom hole pressure, 157, 165 crack-surface displacement, 25
Boussinesq equation, 187, 188 critical crack length, 24
brittle failure, 14, 20, 124 critical pore pressure; criticality, 129, 150–152, 159
brittle–ductile transition zone, 119 critical pore-pressure perturbation, 123
Brown–Korringa equation, 61 critical stress, 24
bulk compressibility criticality of rocks, 150, 154, 155
of grain material, 57 crustal faults, 124
of drained rocks, skeleton, 57, 104, 109 crystalline rocks, 118, 154
of fluid, 56, 177
of rock, 9, 60, 103 Darcy’s law, 63, 76, 184, 200
272
Index 273
decay rate of seismicity, 206 error function, 85

deformation, 1–4, 9, 10, 54, 56, 60, 63, 90 Eulerian formulation, 2
deformation of undrained rock, 90 event probability, 159, 203, 231, 258
differential isochrone, 42 event rate, 159, 203, 206, 220, 258
differential pressure, 103, 104, 108, 114 event-location error, 160
differential stress, 17, 119 exponential diffusion, 190–194
diffusion extensional tectonic regime, 100
equation, 74–81, 84–88, 126, 144, 151, 153, 159 external surface of a porous sample, 49, 51–53
equation, non-linear, 196
front, 199 factorized anisotropy and non-linearity, 197
wave, 70, 74, 128, 138 failure criterion stress, FCS, 20, 99, 124, 131
diffusivity matrix, 136 far field, 13, 33, 92
diffusivity, hydraulic far-field approximation of Green’s function, 33
exponential, 190–192 fault plane, 16, 17, 20, 29
of fracture, 177 fault system, 120, 122
of rocks, 70, 75, 126, 128, 130, 134, 143, 144, 149, fault-displacement vector, 28
151, 152, 161, 166, 172, 177, 185, 189, 212 fault-guided induced seismicity, 122
power law, 182–185 Fenton Hill case study, 157
principal components, tensor, 136 Fermat’s principle, 140
tensor, 134, 135, 141, 154, 155, 157, 161, 162, 196 filtration velocity, 62–64, 144, 166, 184, 200
dilatation, 8, 14, 58, 63, 65, 113 filtration, non-steady-state, 146
dilatational waves, 65–74 fluid flow, 73
dimensional analysis, 186, 190 fluid incompressible, 165
dip, fault plane, 29 fluid viscosity, 63, 64, 126, 136, 177
Dirac function, 12 fluid-loss coefficient, 166, 172, 173, 176
directivity effect, rupture propagation, 35 fluid-mass conservation, 63, 186, 203
Dirichlet boundary condition, 190 fluid-mass diffusion, 74
dispersion equation, 11, 68 fluid-mass dipole, 82
displacement, 1, 10, 13, 14, 49–51, 53, 62–64, 66, 67, focal mechanisms, 36
71, 74, 77, 80, 82–86, 90, 92, 93 foot wall, fault, 27, 28
double couple, 32 fractal, 238, 239
double-dot product, 5, 8, 58, 61 fracture closure pressure, 165
dynamic features of induced seismicity, 159 fracture toughness, 26
dynamic friction coefficient, 21 Frenkel–Biot equations, 124, 125
dynamic viscosity, 63 Frenkel–Biot slow wave, ix
friction angle, 19, 20, 131
earthquake amplitude spectrum, 35 friction coefficient, 17–20, 234
earthquakes, 14 coefficient of internal friction, 17
effective diffusivity, 143, 145, 190, 195
effective medium, 105, 107, 143, 154 gas shale, 166, 177–179, 244
effective normal stress, 124 Gassmann equation, 61, 103
effective permeability, 144, 177 geometric non-linearity, 101
effective stress, 18, 50, 52, 54, 58, 103, 133 geometrical spreading, 131, 133
effective-stress coefficient, 57, 103, 113–115 geometrical-optics approximation, 138, 142, 143
effective-stress coefficient for porosity, 103 geothermal site, 154, 209, 219, 253
effective-medium theory, 105 geothermal system, 89, 154, 180, 219, 242, 244
eikonal equation, 138–143, 146 German Continental Deep Borehole, KTB, see KTB
elastic anisotropy, 10, 126 global flow, 62, 68, 73, 76
elastic body, 2 grain material, 49, 50, 53, 56–61, 102, 110, 111
elastic compliances, 5, 7, 48 Green’s function, 12, 13, 31, 81, 83, 84, 148
elastic deformations, 2 Griffith’s failure criterion, 22, 23, 26
elastic forces, 2–4, 10, 64, 73 group velocity, 12
elastic moduli, 6, 8, 108–110 Gutenberg–Richter law, 216, 222, 224, 227, 238, 240,
elastic piezo-sensitivity, 109 242, 244
elastic stiffnesses, 5, 7, 8 Gutenberg–Richter statistics, 205, 230
elastic strain energy, 5, 6
elastic waves, 10–13, 30–35, 65, 73, 125 hanging wall, fault, 28
ergodicity assumption, 147, 150 harmonic function, 14
274 Index
Hashin–Strikman bounds, 105, 106, 110 moment, seismic

Haskell fault model, 35 density tensor, 31
Heaviside function, 12, 83, 147 magnitude, 36
Hooke’s law, 4–6, 8–10, 51–53, 55–57, 62, 86, 102 rate, 34
hydraulic fracture, 149, 164–177 scalar, 36
hydraulic fracturing, x, 165, 169, 176, 177, 181, 182, spectrum, 34, 35
187, 188, 219, 242, 244, 253 tensor, 33
hypocenter, 42, 123, 129 momentum conservation, 10, 62
monoclinic media, 7
index of non-linearity, 185, 198
induced and triggered events, 250 Neumann boundary condition, 199
injection pressure, 118, 155 non-linear diffusion of pore pressure, 163, 177–200
inter-event time, 220 non-linear elastic moduli, 110
internal friction, 17, 20 non-linear fluid-rock interaction, 203
internal surface of a porous sample, 49, 51–53 normal elastic waves, 94
irrotational displacement field, 74, 75, 80, 82, 83, 90, normal fault, 29
93 normal faulting, 15, 100
isothermal elastic moduli, 9, 10 normal stress, 16
isotropic medium, 8–11, 13, 33–37, 57–62, 151
Ogachi case study, 209
jacketed sample, 52 Omori law, 206
optimal angle, fault plane orientation, 19, 20
kinematic features of induced seismicity, 149, 159 orthorhombic media, 7, 60
Kozeny–Carman relation, 114
Kronecker matrix, 6 P-wave modulus, 12, 66
KTB, 40, 43, 107, 109, 118–124, 156, 190, 218, 219, Paradox Valley case study, 210
245 peak ground acceleration, 220
penny-shaped crack, 23, 24, 105, 107, 229
Lagrangian formulation, 1, 2 permeability, hydraulic
Lamé equation, 10 compliance, 116, 190
Laplace equation, 14 of rock, ix, x, 10, 114, 122, 126, 136, 149, 162,
Laplace operator, 14 172, 177, 185
linear diffusion, 133 tensor, 63, 64, 126, 134, 137, 184
linear elasticity, 1–13, 25–33, 53, 83 upscaling, 134, 145
linear relaxation, ix phase function, 133
linear vector dipole, 32 phase slowness, 140
local flow, 72 phase velocity, 11, 69–72
location error, 44, 133, 154, 251 piezo-sensitivity, 109–112
lock-up angle, 19 -theorem, 186
longitudinal (P-) waves, 11, 12, 42, 69, 72, 78, 93, 125 PKN model of a hydraulic fracture, 166
plane wave, 10, 11, 67, 70, 78, 146
magnitude probabilities, 202, 205, 253 point source of fluid injection, 83, 159
Mandel–Cryer effect, 93 Poiseuille flow, 64
material infilling pore space, 54, 55 Poisson process
maximum compressional stress, 96–100, 165 homogeneous, 220
maximum induced magnitude, 252 non-homogeneous, 221
mean stress, 17 Poisson’s ratio, 9, 59, 86, 110, 112
mesoscopic flow, 73 polarization, 11, 42, 68, 71
microearthquakes, 37 pore pressure
Microseismic monitoring, ix diffusion, ix, 74, 149, 153, 168
microseismicity triggering, ix front, 189
minimum compressional stress, 96, 165 in rock, 18, 49, 102, 113, 126
minimum critical pressure, 207 perturbation, time harmonic, 126, 127, 139, 150
minimum-compression in-situ stress, 97 relaxation, 168
modes of cracks, 24 pore space, 48
modified Griffith failure criterion, 23 porodynamics, ix
Mohr’s circle, 17, 22, 98 poroelastic coupling, 94–113, 134
Mohr–Coulomb failure criterion, 18–23, 227, 239 poroelastic stress coefficient, 85, 86, 112, 131
Index 275
poroelastic systems, 48–52 shear (S-) waves, 11, 42, 66, 69

porosity, 49, 53, 102–114 shear crack, 24
potency density tensor, 28, 29 shear failure, 14
power-law, 116, 185, 204, 208 shear modulus, 9, 62, 106
power-law diffusion, 185, 192 shear stress, 17–19
power-law diffusivity model, 191–196 Skempton coefficient, 59, 61, 113
power-law function, 114, 238 slip event, 28, 166
power-law statistics, 229 slip law, 21
principal tectonic stresses, 15, 97, 118 slip vector, 28–30
probability density function, criticality, 150–153, 155, slow wave, 72–94, 125, 128, 140, 146
157, 159 slow wave, harmonic spherical, 139
probability density function, PDF, 150, 221, 230, slow wavefield, 72–94, 143
237–239 slowness, 139
probability of earthquake occurrence, 159, 202, 203 small deformations, 2
probability of seismic events, 159, 201–224 Soultz case study, 154, 207, 221
proppant, 168, 172, 219 spatial averaging, 144
spatial density of events, 159
quasi P-waves, 11, 45 spatio-temporal evolution of seismicity, 125, 127–130
quasi S-waves, 11, 45 specific magnitude, 217
quasi-static approximation, 76–95 spherical wave, 13, 33, 133
squirt flow, 73
r –t plot, 128–132, 165, 175, 178, 190 stability condition, 6, 9, 44
radial stress, 92, 93 statistical ensemble, 147, 150
rake, fault plane, 29 statistically homogeneous random field, 202
random criticality field, 150 statistics of criticality, 150
reciprocity theorem, 54, 116 steady-state filtration, 145
reference seismicity rate, 207 step function, 127
relaxation radius, 128, 191 stiff porosity, 105, 110, 111, 115
relaxation zone of pore pressure, 128 stiffness tensor of skeleton, 57
representative volume, 48, 62, 64 stiffness tensor of undrained rock, 62
reverse fault, 29 stimulated volume, 224, 227–229, 232, 235
reverse faulting, 100 storage coefficient, 70, 126, 144, 200
Rhine Graben, 154 strain tensor, 1, 2, 10, 27, 50, 54, 113
rigid motions, 1 strength of injection source, 92, 138, 149, 151, 155
rock failure, 20, 22, 113, 114, 150 strength of pre-existing cracks, 150
rotational moments, 3, 32 stress drop, 26, 27, 36, 237–248, 252
rotational waves, 65, 67, 68, 71 stress force, 2, 25
rupture surface, 27, 35, 202, 224, 227–238 stress intensity factor, 25
rupture velocity, 26, 35 stress sign notation, 15
stress tensor, 3, 4, 13, 51, 57, 60, 78, 86
saturating fluid, 48, 53 stress–strain relations, 4, 49, 55, 59, 87
scalar hydraulic diffusivity, 128, 154 strike, fault plane, 29
second dilatational wave, 70–72 strike-slip fault, 29
second Newtonian law, 10, 62 strike-slip faulting, 15
seismic emission, 37 subduction zone, 235
seismic event, 1, 127 summation on repeated indices, 4
seismic hazard indicator, xi, 220 superposition principle, 12
seismic moment, 36, 215, 237 surface energy per unit area of a crack, 24
seismic multiplets, 38
seismic P-wave, 12, 69, 76 tectonic potential, 204, 211
seismic reflectors, 120, 122 tectonic stress, 15, 30, 95
seismic trace, 37–40 tensile crack, 22, 24
seismic velocity, 106, 108 tensile stress, 4, 22
seismic waves, 10, 32, 37, 59, 72, 73, 75, 82, 83 Terzaghi’s effective stress, 18, 114
seismicity triggering, 118 thrust fault, 29
seismogenic index, 201, 215–222 thrust faulting, 15, 100
seismograms, 37 tight gas reservoir, 166, 169
self similarity, 238 tight gas sandstone, 166–173
276 Index
tight rock, 9, 109, 111, 113, 132, 166 vector of fluid flux, 62, 146
time-harmonic plane wave, 11, 67 velocity strengthening, 21
torques, 3 velocity weakening, 21
tortuosity, 65 viscosity of a fluid, 63, 64, 145, 177
total moment tensor, 33 Voigt’s notations, 6
total stress, 49, 86
volume averaged strain, 27, 50, 53
traction, 2–4, 15, 16, 49, 95, 117
volume balance, 164, 165, 168, 181, 187
transformation strain, 30
transverse isotropy, 8, 43–45, 60 volume of a hydraulic fracture, 166
treatment fluid, 165, 166, 168, 172, 177, 178 volumetric hydraulic fracturing, 177, 180, 183
triclinic media, 7 volumetric strain, 8
triggering front, 118, 127–149, 152, 158, 161–166,
176–200 wave number, 68–71, 73, 142
undrained rock, 55, 59–62, 66, 69, 83, 88–90, 94, 108 wave vector, 11, 67, 68, 70, 71
undrained system, 54, 55, 59, 61, 79 wavelength, 13, 32, 33, 71, 128
uniaxial strain, 87 weak anisotropy, 44
uniaxial tensile strength, 22, 165 well-head pressure, 119, 120, 155, 157
unjacketed sample, 50, 52 width of fracture, 173
27.05.04 01.06.05
10 160
9 (a) 140
8
120
# Events/Day
7
6 100
5 Surface stations 80
4 60
3
40
2
1 20
0 0
0 25 50 75 100 125 150 175 200 225 250 275 300 325 350
140 (b) 2800
120 2400
# Events/Day
100 3485 m 2000

80 1600
60 Main-hole sonde 3500 m 1200
No data from
40 sonde 800
20 1950m 400
0 0
0 25 50 75 100 125 150 175 200 225 250 275 300 325 350
S-P travel time at HBR [s]
(c)
0.3 0.3
0.2 0.2
0.1 0.1
0.0 0.0
0 25 50 75 100 125 150 175 200 225 250 275 300 325 350
100 000
Injection Volume [m3]

140 (d)
Volume
Pressure [bar]
80 000
120 Pressure 60 000
100
40 000
80
20 000
60
0
0 25 50 75 100 125 150 175 200 225 250 275 300 325 350
Days since start of injection experiment (27/05/04)
Plate 3.1 Microseismicity during the 2004–2005 KTB injection experiment.

(a) Events recorded by the surface stations. (b) Events recorded by the geophone
located in the main borehole. (c) Time differences between arrivals of S-and
P-waves. Changes in the locations of the borehole geophone are clearly seen
from the plot. (d) The injection pressure along with the cumulative volume of
the injected water. The time when the amount of previously extracted fluid was
re-injected is marked in part (a) of the figure by the inverse triangle on the top,
between the days 100 and 125. Another triangle (between the days 325 and 350)
marks the injection termination. (Modified after Shapiro et al., 2006a.)
Plate 3.2 A vertical slice of a 3D depth migrated image of the KTB site (see
Buske, 1999). Light tones correspond to high seismic reflection intensities and
dark tones to lower ones, respectively. The SE1 reflector is clearly visible as a
steeply dipping structure. Additionally, seismicity induced by the injection exper-
iments of years 1994 (close to the injection depth of 9.1 km), 2000 (injection
depth of approximately 5.6 km) and 2004–2005 (depth around 4 km) is shown.
Locations of the main (black line) and pilot (light-tone line) boreholes are also
plotted. (Modified after Shapiro et al., 2006a.)
Plate 3.3 A horizontal slice at 4 km depth over the depth-migrated 3D image of

the KTB site (Buske, 1999) plotted along with the slice-plane projections of seis-
mic hypocenters induced by the injection experiment of 2004–2005. The white
and gray squares are locations of the main and pilot boreholes, respectively. (After
Shapiro et al., 2006a.)
50 50
40 40
30 30
20 20
10 10
0 0
–10 –10
–20 –20
–30 –30
–40 –40
–50 –50
–50 –40 –30 –20 –10 0 10 20 30 40 50 –50 –40 –30 –20 –10 0 10 20 30 40 50
Plate 3.6 Left: realizations of the critical pore pressure C randomly distributed in
space with an exponential auto-correlation function. Right: the same but with a
Gaussian auto-correlation function. (Modified from Rothert and Shapiro, 2003.)
Events and occurence time Events and occurence time

50 50 100
40 40 90
30 30 80
20 20 70
Distance [m]
Distance [m]
10 10 60
0 0 50
–10 –10 40
–20 –20 30
–30 –30 20
–40 –40 10
–50 –50
–50 –40 –30 –20 –10 0 10 20 30 40 50 –50 –40 –30 –20 –10 0 10 20 30 40 50
Plate 3.7 Synthetic microseismicity simulated for the two spatial distributions
of the critical pore pressure shown in Figure 3.6, respectively. (Modified from
Rothert and Shapiro, 2003.)
1.2
Reference rate R0
1
Cmin = 1000 Pa
0.8 Cmin = 10 000 Pa
Rs(t)
Omori law (p = 1.8)

0.6
Omori law (p = 3)
0.4 Omori law (p = 10)
0.2
0
0 1000 2000 3000 4000 5000 6000 7000 8000 9000
t[s]
Plate 5.1 Seismicity rate in the case of stable pre-existing fractures (Cmin is finite
and significant). The rate is normalized to the reference seismicity rate ν I defined
in (5.17). Parameters of the model are D = 1 m2 /s, t0 = 2000 s, p0 = 1 MPa,
a0 = 4 m, Cmax = 1 MPa. The solid lines correspond to Cmin = 0 Pa,
Cmin = 1000 Pa and Cmin = 10 000 Pa (from the upper to the lower curve).
The dashed lines show the modified Omori law with pd = 1.8, 3.0 and 10.0.
(After Langenbruch and Shapiro, 2010.)

1.2
Reference rate R0
1
Cmax = 105 Pa
0.8
Cmax = 104.5 Pa
Rs(t)
0.6 Cmax = 104 Pa

0.4
0.2
0
0 1000 2000 3000 4000 5000 6000 7000 8000 9000
t[s]
Plate 5.2 Seismicity rate in the case of unstable pre-existing fractures (Cmin = 0).
The rate is normalized to its maximum value, given by the reference seismicity
rate ν I . Parameters of the model are D = 1 m2 /s, t0 = 2000 s, p0 = 1 MPa,
a0 = 4 m. The solid lines correspond to Cmax = 106 Pa, Cmax = 105 Pa,
Cmax = 104.5 Pa and Cmax = 104 Pa (from the upper to the lower curve). The
arrows denote the time of maximum probability to induce an event with significant
magnitude. (After Langenbruch and Shapiro, 2010.)
Ogachi 1991
104
flow rate
pressure
103 M > –2.5
M > –2.0
NE VM>M0
M > –1.5
102
101
1 5 10 30
Time (days)
Plate 5.3 N≥M as functions of injection time for the Ogachi 1991 experiment. The
points are observed cumulative numbers of earthquakes with magnitudes larger
than the indicated ones. The straight line has the proportionality coefficient 1,
predicted by equation (5.14). The curves show the injection pressure (the lower
line in the time range 1–10 days) and the injection rate (the upper line in the time
range 1–10 days). (Modified from Shapiro et al., 2007.)
104
Ogachi 1991 –2.5
Ogachi 1993
Paradox Valley 1996–2004 –2.0
–1.5
103
NEVM >M0
–1.0
–0.5 0.5
2 1.0
10
1.5
2.0
1
10 –1
100 101 102 103 4
10 10
Time (days)
Plate 5.5 A combined plot of numbers of events with magnitudes larger then
given ones as functions of injection durations at Ogachi and at Paradox Valley.
Thin dashed lines correspond to equation (5.14) with i = 0. (Modified from
Shapiro et al., 2007.)
2
1
0 2
3
4
−2 5
6
7
8
−4
Σ
9
10
11
−6 12
13
14
−8 15
16
−10
0 0.2 0.4 0.6 0.8 1
Time/Injection Time
Plate 5.10 Seismogenic index computed for different locations of Enhanced

Geothermal Systems, hydraulic fracturing in hydrocarbon reservoirs, and other
injection locations (injection times are given in parentheses). 1: Ogachi 1991 (11
days), 2: Ogachi 1993 (16 days), 3: Cooper Basin 2003 (9 days), 4: Basel 2006
(5.5 days), 5: Paradox Valley (2500 days), 6-9: Soultz 1996 (48 hr), 1995 (11
days), 1993 (16 days) and 2000 (6 days), 10: KTB 2004/05 (194 days), 11–12:
KTB 1994 (9 hr) [upper and lower bound, calculated for two b-values], 13: Bar-
nett Shale (6 hr), 14–16: Cotton Valley Stages A (2.5 h), B (2.5 h) and C (3.5 h)
(Modified from Dinske and Shapiro, 2013.)
x
250 m y
0.001 500 m
z
0.01
0.1
1
1000 m
D (m2/s)
Plate 3.13 Reconstruction of the diffusivity distribution in a hydraulically hetero-

geneous geothermic reservoir of Soultz (corresponding to the stimulation of 1993)
using the eikonal-equation approach. The dark-tone grid cells in the upper part of
the structure have hydraulic diffusivity in the range 0.1–1 m2 /s. Below 1000 m
the structure has the diffusivity mainly in the range 0.001–0.05 m2 /s. (Modified
from Shapiro et al., 2002.)
100
10–1
10–2
PDF Δτ
Soultz 93
Soultz 93
10–3 Soultz 94
Soultz 95
10–4 Soultz 96
Soultz 2000
Basel
10–5 HPP
exp(−Δτ)
10–6
10−2 10−1 100 101
Δτ
Plate 5.11 Estimated probability density functions of the normalized inter-event

time for stationary periods of injections at Soultz and Basel. (After Langenbruch
et al., 2011.)
100 injected volume
M > 0.5 (Mc)
M>1
10–1 M > 1.5
M>2
10–2
10–3
100 101 102

Time since first event (h) shut-in time
Plate 5.13 Number N≥M (solid lines) of induced earthquakes with magnitudes
M larger than indicated values as functions of the time t elapsed from the time
of the first event in the catalog (nearly the injection start) at the Basel borehole
(Häring et al., 2008). The plot also shows the injected water volume Q c (t) (the
upper dashed line). The quantities Q c (t) and N≥0.5 (t) (the upper solid line; 0.5
is approximately a completeness magnitude) are normalized to their values at the
moment of the maximum injection pressure (several hours before the injection
termination). Immediately after the injection termination the curve Q c (t) starts to
decrease because of an outflow of the injected water. Theoretical curves N≥M (t)
corresponding to (5.20) are given by lower dashed lines. They are constructed by
a time-independent shifting of the curve of the injected-fluid volume. The lower
solid lines shows the observed quantities N≥M (t) normalized by the same value
as the quantity N≥0.5 (t). (Modified from Shapiro et al., 2011.)
8 8
lower bounds lower bounds
upper bounds upper bounds
7 Φ = 0° 7 Φ = 0°
Φ = 5° Φ = 5°
6 Φ = 10° 6 Φ = 10°
Φ = 15° Φ = 15°
Φ = 20° Φ = 20°
5 Φ = 25° 5 Φ = 25°
Φ = 30° Φ = 30°
log10 Nev
log10 Nev
Φ = 35° Φ = 35°
4 Φ = 40° 4 Φ = 40°
Φ = 45° Φ = 45°
3 Gutenberg–Richter 3 Gutenberg–Richter
2 2
1 1
0 0
−2 −1 0 1 2 3 4 5 −2 −1 0 1 2 3 4 5
Magnitude Magnitude
Plate 5.21 The same as Figure 5.20 but cuboidal stimulated volumes and different
angles φ. (After Shapiro et al., 2013.)

Fluid Induced Seismicity

Uploaded by

Copyright:

Available Formats

Fluid Induced Seismicity

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Fluid Induced Seismicity

Uploaded by

Copyright:

Available Formats

FLUID-INDUCED SEISMICITY

The characterization of fluid-transport properties of rocks is one of the most impor-

SERGE A. SHAPIRO is Professor of Geophysics at the Freie Universität Berlin,

Cambridge University Press is part of the University of Cambridge.

2.2.1 Independent confining stress and pore pressure 55

3 Seismicity and linear diffusion of pore pressure 118

3.8 Strength of pre-existing fractures 150

5.5 Appendix 3. Probability of discs with centers inside a sphere 256

Color plate section between pages 82 and 83

Characterization of fluid-transport properties of rocks is one of the most important,

related to slow waves in porous fluid-saturated materials. The dynamics of slow

This book contains results of research funded by different institutions in differ-

By “seismic events” we understand earthquakes of any size. There exists a broad

1.1 Linear elasticity and seismic waves

Figure 1.1 A sketch for defining a traction.

1.1.3 Stress–strain relations

Figure 1.2 A sketch illustrating positiveness of tensile stresses. Indeed,

i j = Si jkl σkl , (1.7)

Si jkl = S jikl = Si jlk . (1.8)

σi j = Ci jkl kl . (1.9)

Ci jkl = Ckli j , Si jkl = Skli j . (1.13)

1.1.4 Elastic moduli

coordinate orientation, three additional constants (corresponding to three rotational

Let us consider a volumetric strain (dilatation) of an elementary volume V of an

Let us further assume that this dilatation is a result of a hydrostatic stress,

The proportionality coefficient here is a bulk compressibility C mt of the elastic

1.1.5 Dynamic equations and elastic waves

In linear systems any wavefield can be decomposed into a superposition of

Here κ is a wave vector of an arbitrary direction and an unknown length and u0 is

Ci jkl u 0k κ j κl = ρω2 u 0i . (1.31)

(Ci jkl κ j κl − ρω2 δik )u 0k = 0. (1.32)

This matrix equation is called the Christoffel equation. This is a homogeneous

det(Ci jkl κ j κl − ρω2 δik ) = 0. (1.33)

1.1.6 Point sources of elastic waves

Sometimes it is convenient to work with source terms having Heaviside-type

1.1.7 Static equilibrium

1.2 Geomechanics of seismic events

1.2.1 Faults and principal stresses

σzz = −σ1 , σ yy = −σ2 , σx x = −σ3 . (1.53)

σzz = −σ2 , σ yy = −σ3 , σx x = −σ1 . (1.54)

σzz = −σ3 , σ yy = −σ2 , σx x = −σ1 . (1.55)

Thus, the traction component τt tangential to the fault plane is

σ3–Pp σ1–Pp σn–Pp

geometric consideration. Recalling that in Figure 1.5 the friction coefficient μ f

Note that an equivalent consideration of the half-space (σn − Pp , τt ) with negative

1.2.2 Friction coefficient

A great deal of active research in modern seismology has been dedicated to

1.2.3 Growth of finite cracks: a sufficient condition

For σc = 0 and T0 = Cc /2 this equation reduces to the Mohr–Coulomb criterion

1.2.4 Necessary conditions of crack growth and some results of

stress τl (this stress must be specified correspondingly to the loading configuration)

Here δ E s = γc δS is an energy increment needed to create a new crack surface δS

where C g is a proportionality coefficient. For different modes of cracks, i.e. a ten-

The left-hand side of equation (1.89) contains a combination of rock physical

1.2.5 Earthquake motions on faults

ij d 3 r = pi j (r)d 2 r, (1.103)

where the quantity

1.3 Elastic wavefields radiated by earthquakes

where the function

where r represents an (averaged) source location and

ij d 3 r = pi j (r)d 2 r, (1.103)

where r represents an (averaged) source location and

Mi j (t) = Ci jkl pkl (t)δ D (c+ )d 3 x̂

where p is the surface of the pore space, r is a point on this surface, u i is a