Documentos de Académico
Documentos de Profesional
Documentos de Cultura
Abstract
Rock heterogeneities exist in mostly all rock types present in the Earths crust. Fractures are
one such heterogeneity, which can range in size from very small micro-scale diffuse fractures to
huge kilometer-wide regional features and meso-scale fracture corridors. The characterisation
of fracture properties has many important implications for a wide range of industries. One
way of characterising fractures is by making use of numerical forward modeling experiments.
By observing how seismic waves are being affected when travelling through fractured media,
information about the fractures might be obtained. In this study, seismic wave propagation through various fracture models is investigated, using a finite difference modelling code
(FDELMODC). The main objectives of this work are to determine whether the effects of
fractures can be seen in the seismic response, and in particular whether the seismic method
can be used to detect fracture networks. In addition to investigating whether the detection of
fractures and fracture networks is possible, this study also focuses on the question of whether
fractures and fracture patterns can be identified. Several fracture characteristics have been
investigated: the effects of fracture length, width and orientation on the recorded reflected
and transmitted signals, as well as the effects on different fracture fill. Starting with observations for single fractures, the models have been slowly extended towards more complex
and realistic scenarios. Since the forward modelling involved the elastodynamic wave system, the effect of fractures on different wave types, namely P waves and S waves, has been
investigated. In this study it was shown that the presence of fractures in a medium lowers
the amplitudes in the frequency domain, regardless of the fracture characteristics or the wave
type being considered. For a single fracture, many results showed a clear orientation effect
and length effect. The analysis between P and S waves led to some very interesting results,
with the different waves proving sensitive to different orientation due to their particle motion
characteristics. When the effect of larger numbers of fractures was investigated, some limitations of the seismic method for fracture characterisation emerged. Although the presence of
fractures seems to significantly lower amplitudes in general, it proved harder to distinguish
between different fracture patterns when large clusters of fractures were considered. When
the orientations and positions of a group of fractures was changed from completely random
to fixed positions very little difference was observed in the resulting data, making it hard to
infer anything about the characteristics. The results of this work suggest that while fractures
can certainly be detected using seismic methods, they cannot really be identified to any great
certainty. In reality, fractures most often occur in clusters and groups. The results of this
study indicate that these groups are much harder to characterise than single fractures or
faults, thereby implying that in realistic scenarios more sophisticated methods are required.
vi
Abstract
Acknowledgements
I would like to thank my supervisors Dr Auke Barnhoorn and Dr Ir Jan Thorbecke for their
help and guidance throughout this project. It was really inspiring to work on this multidisciplinary topic. Their comments on the manuscript have improved the work significantly.
Thanks for everything that I have learned throughout the process of this master thesis. It
has been a great experience of doing research.
viii
Acknowledgements
Table of Contents
Abstract
Acknowledgements
vii
. . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3 Numerical modelling
10
11
3-3-1
11
3-3-2
11
3-3-3
12
13
15
15
4-2
16
17
18
ix
Table of Contents
21
21
23
23
27
28
Frequency domain . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
29
31
43
5-3-1
45
model definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
frequency domain observations . . . . . . . . . . . . . . . . . . . . . . .
45
45
45
.
.
.
.
48
48
48
48
50
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
model definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
time domain observations . . . . . . . . . . . . . . . . . . . . . . . . . .
frequency domain observations . . . . . . . . . . . . . . . . . . . . . . .
50
51
52
7 Discussion of results
7-1 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
7-2 Future recommendations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
53
57
58
6-3-1
6-3-2
6-3-3
Bibliography
59
Chapter 1
In addition to the advancements in the seismic acquisition stage, utilization of the forward
modelling method is gaining popularity for fracture characterisation. Though it has long
been recognised as a powerful tool for understanding wave propagation in general. With the
growing interest of seismic wave propagation in fractured rocks, more studies are emerging
that aim to characterize the elastic wave propagation behavior using this technique. Among
the many numerical forward modelling techniques available, Finite difference modeling is one
such method that is relatively easy to understand and implement. Other numerical techniques
available include the finite element method and the pseudo spectral method.
Given this motivation of characterising fractures using the finite difference method, the aim
of this study is to use the programme FDELMODC to model the seismic wave propagation
through various fracture models. The main objectives of this work are to determine whether
the effects of fractures can be seen in the seismic response, and in particular whether the
seismic method can be used to detect fracture networks. In addition to investigating whether
the detection of fractures and fracture networks is possible, this study tries to answer the
question of whether fractures and fracture patterns can be identified. In order to examine the
effect of fractures on the seismic signal, results are discussed in both the time and frequency
domains.
This thesis begins with a recapitulation of the theory of fractures and seismic wave propgation. Following this, the concept of finite difference modelling is described and the modelling
procedure for this study is explained. The results of this report can be fit into three catergories, those related to the modelling grid (can be found in chapter 4), those related to a
single fracture (chapter 5) and those related to more complex models (chapter 6). These
results will be evaluated and interpreted separately in their individual chapters and discussed
all together in the discussion section found in chapter 7.
Chapter 2
Most rocks contain fractures of some kind as well as other types of heterogeneities including
faults, foliation and geological discontinuities. Despite this, many aspects of fractures such
as their permeability and porosity characteristics are poorly understood.
The term fracture can in reality be used to describe a wide range of features of varying scales,
including larger scale faults, gashes, veins, fissures etc, making a precise definition difficult.
In a general sense a fracture is recognised as being a discontinuity or a plane of weakness in
the rock, which when put under stress causes a partial loss of cohesion in the rock. More
specifically a fracture can further be used to describe a break in a rock without any or little
displacement parallel to the surface of the fracture. In this sense the definition is narrower
and does not include more distinct features such as faults. This definition of a fracture may
also be described as a joint [Twiss and Moores, 1992].
Fractures tend to form in clusters, for example the term fracture zone describes highly interconnected discrete fractures which are spaced closely together, while a fracture corridor refers
to tabular fracture groups. An individual fracture has a limited spatial extent while fracture zones form a quasi continuous fracture which can be many kilometers in length. Many
adjacent fractures can form sets, and then further the interconnection of numerous sets can
form a network which facilitates fluid flow and increases hydraulic conductivity in a rock. As
such, the characterisation of fractures is concerned with the network of fractures in a rock
rather than any individual feature. These sets and networks can have an extent of hundreds
of kilometers and display a near constant orientation of fractures, except when associated
with folding. In order to build up realistic geological models it is necessary to understand not
only how the fracture network may be distributed, but also the distribution of the fracture
with regards to their orientation and shape [Singhal and Gupta, 1999].
Systematic fractures, those which are planar with a regular distribution, may be classified in
different ways. A geometric classification describes the relationship between the fracture and
the natural bedding of a rock. The order of the fractures relates to their extent through rock
layers and finally the genetic classification refers to the relative motion across the fracture
surface during formation. Besides systematic fractures, irregular and curved nonsystematic
3
fractures also occur, although normally in combination with older systematic fractures. Individual fractures are typically between a few centimeters and some meters in length, although
extremes may exist on both the milometer and kilometer scale. The spacing between fractures is dependent on the bed thickness and rock type, however fractures normally display a
consistent spacing behaviour. Fractures may also intersect, the angle of intersection is also
normally a predictable rather than a random behaviour. For example the angle of intersection
for small scale shear fractures in always around 60 degrees [Twiss and Moores, 1992].
In order to analyse how fractures may form it is necessary to examine the stress conditions under which they are generated. By considering the Cauchy formula, which relates the traction
across any surface within a continuous body with the stress tensor at that point, in combination with the conditions of equilibrium of forces, and the stress tensor itself, it is demonstrated
that stress can be resolved into normal and shear components. The shear stress is related
to the formation of fractures. In particular fractures will occur along planes where the shear
stress is maximum [Ranalli, 1995].
[Bakulin et al., 2000] justifies the use of the finite difference method for the modelling of
fractures, since the fracture openings are small. They stated that when building effective
seismological models, the long wavelength approximation can be considered, meaning that
fracture openings can be neglected and the details of the spatial distributions of fractures is
also neglected.
As a result, the medium containing the fractures can be considered as an effective anisotropic
solid, where the orientation and intensity of fracture sets, along with the elastic coefficients
of the host rock and fracture fill, will dictate the parameters of the model. This assumption
can be made since seismic wavelengths are many tens or hundreds of meters in length and
hence:
seismic wavelength fracture spacing fracture opening.
With these basic conclusions on the length, spacing, distribution and other fracture characteristics a good base is presented for constraining future geologically complex models, this is
particularly relevant for the results of chapter 6. Additionally the use of the finite difference
method is justified for the modelling of fractures. The principle of seismic wave propagation
will now be discussed in more general terms as well as the behaviour of seismic waves at
fractures.
2-2
In seismic surveying a seismic source is used to generate a pulse which propagates away from
the source as a spherical wavefront. There are two kinds of seismic waves, those that travel
through the elastic solid known as body waves and those that propagate along the boundary
of the solid known as surface waves. The latter will not be considered in this study.
Concerning the body waves, a further distinction can be made depending on their mode of
strain. Compressional P waves propagate by compressional and dilational uniaxial strains
in the direction of wave travel. Individual particles also oscillate in the direction of travel.
Shear S waves propagate by a pure shear strain perpendicular to the direction of wave travel.
Hence, individual particles also oscillate in this direction at right angles to the direction
of wave travel [Kearey et al., 2002]. The mechanical disturbance caused by the waves as it
propagates results in a counter force which tries to restore the equilibrium situation. If the
medium is liquid it can only support a compressional restoring force, and as such only P waves
can propagate in liquids. Solids however, can support both compressional and shear restoring
forces and both body waves can propagate through [Wapenaar and Berkhout, 1989].
The velocity at which these P and S waves travel through the medium is determined by the
elastic moduli and the density of the material. Since the P and S waves involve different strains
in their propagation, their elastic modulus expressions are different, resulting in different
velocities. Equations 2-1 and 2-2 show the different expressions for the velocities of P (vp )
and S (vs ) waves
vp
vs
"
K + 43
=
1/2
=
,
#1/2
(2-1)
(2-2)
respectively, where K is the bulk modulus, represents the shear modulus and denotes the
density.
For any given medium, the two expressions are related not only to the medium properties but
to each other. The poissons ratio () of the rock is linked to its lithology, and determines the
ratio between the P and S propagation velocity, as shown in equation 2-3.
2(1 ) 1/2
vp
=
vs
(1 2)
(2-3)
Poissons ratio is known for most rocks, for an consolidated rock mass the value is typically
taken as 0.25. Substituting this value into equation 2-3 gives a ratio of vp=1.7vs. Given this
equation, it follows that P wave velocity is always faster than the S wave velocity for the same
medium. As different rock types display different densities and elastic moduli, it follows from
equations 2-1 and 2-2 that their P and S wave velocities differ.
The P and S body waves have associated with them a frequency, For nondispersive waves all
frequency components are determined by the same elastic moduli and density and so hence
travel at the same velocity. This behaviour explains that the velocity of the propagating wave
is independent of the frequency of the wave [Kearey et al., 2002].
The importance of seismic waves for imaging in the subsurface, lies in their behaviour at layer
boundaries. As described above, the wavefield is emitted from the source and propagates
downwards through the subsurface until it reaches a boundary where different behaviours are
then observed. In particular, waves are reflected back to the surface or transmitted through
the boundary. The total measured seismic response at the surface can be described by the
two-way wave equation, which describes downward and upward wavefield propagation, as well
as reflections. Here it is described following the approach by [Wapenaar and Berkhout, 1989].
To arrive at this equation for the acoustic case the laws of conservation for mass (equation
2-4) and momentum (equation 2-5) need to be considered.
~v dV =
dV = ~v .~ndS +
t
S
V
im dV
(2-4)
~ (V )
(~v )~v .~ndS + F
(2-5)
where V is the volume, is the volume density of mass, S is the surface, ~n is the normal vector
pointing outwards from the surface, ~v is the particle velocity, im is the source distribution and
F~ is the resultant force acting on the particles. Applying Guass theorem to both equations,
results in the non linear equations of continuity (equation 2-6) and motion (equation 2-7)
respectively. Here p denotes the pressure and f is the volume density of the external force.
im
+ .(~v ) =
t
t
(2-6)
(~v )
+ ~v .(~v ) + (~v .)~v + p = f~
t
(2-7)
(2-8)
(2-9)
Where K~r is the adiabatic compression modulus, p is the acoustic pressure, ~v is the particles
velocity, (~r) is the static mass density. Finally iv is the volume density of volume injection,
and f~ is the volume density of the external force.
Finally by eliminating particle velocity from the set of equations above, the 2 way wave
equation is derived (equation 2-10).
.
1~
1
2p
2 iv
p
= 2 .
f
2
K t
t
(2-10)
In forward modelling, where seismic data is simulated based on the defined seismic source and
a model of the subsurface, the two-way wave equation is applicable. Here the acoustic twoway wave equation has been considered, normally reserved for an inhomogeneous ideal fluid.
Therefore an approximation has been made that the Earths subsurface is an ideal fluid which
supports P wave propagation and acquisition only. When considering S wave propagation,
this approximation doesnt hold and the elastic case for waves in solids needs to be taken.
As would be expected, the expression for wave propagation in a solid is more complicated
than that in a fluid since an additional shear force is supported. For the most simple case
of a homogeneous isotropic solid , P wave and S wave propagation is independent, with the
wave motion of the P waves being curl free and the wave motion of the S waves divergence
free. However, for a more realistic case quasi-compressional P and S waves are interrelated.
To arrive to the wave equation for an elastic case (equation 2-11), the law of conservation of
momentum needs to be considered again but with a different traction term, which acts on the
surface rather than its normal. Furthermore, the full stress and strain tensors are considered.
These tensors are mutually dependent resulting in a lineared stress-strain relationship, a
generalisation of Hooke law. Substitution of the strain tensor elements into this relationship
gives the stress displacement relation, and by considering the particle velocity vector, the
linearized equation of continuity or stress velocity relation is derived. By eliminating the
stress tensor components or the traction, the 2 way wave equation is derived for an elastic
medium [Wapenaar and Berkhout, 1989].
j (Cjl l~v ) t 2~v = t (f~ j ~j )
(2-11)
Here, Cjl are the stiffness subtensors,~v is the particle velocity vector, is the volume density
of mass, f~ is the volume density of external force and j represents the stress distribution
associated with the source.
So far, the total seismic response has been described which includes the reflection response.
To understand the effect of a fracture, which is assumed to be a very thin layer, it is necessary
to understand what happens at a seismic interface. At such a boundary is normally a change
in velocity and density due to the difference in material. The seismic wavefield will split into
waves which are reflected and transmitted. The amount of energy which is reflected relative
to the amount transmitted relies on the acoustic impedance (Z) contrast at the boundary, as
shown in equation 2-12, this is the difference in the product of the velocities (v) and densities
() of the layers. The ratio between the ray incident on the the boundary and that which is
reflected and transmitted gives the reflection and transmission co-effficient respectively.
Z = v
(2-12)
The velocity of the seismic wave through the rock and the depth of the rock determine the
seismic traveltime of the wave. Adapting this expression for both downward and upward
propagation from source to receiver results in the well known expression for two-way travel
time of the seismic wave.
The seismic response at the surface is measured as a seismic trace representing the impulse
response of the ground. This trace can be thought of as the convolution of the initial seismic
source wavelet with a time series. This time series is a combination of reflection coefficent
and associated two way travel time for each reflection boandary encountered.
Mostly, only P waves are used for exploration purposes, since the data is cheaper and easier
to acquire and interpret. A receiver is typically used that is sensitive only to the vertical
direction so that only the faster longitudinal P wave is recorded. In recent times, three
component recording is becoming more common and the recording of S waves increasingly
important. This is due to the fact that the analysis of S waves may bring much additional
information in seismic surveying. For example [Kearey et al., 2002] states that S waves have
a great potential for detecting gas-filled reservoirs, since the presence of gas would have the
effect of lowering the elastic moduli value and Vp/Vs ratio.
In this study, the seismic response of S waves will also be included in addition to P waves since
it is clear they may bring additional information of importance for fracture characterisation.
As stated previously, P and S waves display different ground motions, vertical and horizontal.
This also means that they might have different sensitivities to different structures such as
fractures. It is also expected that the different waves can give information about the fracture
such as the fracture fill.
2-3
In the section above, the behaviour of seismic waves at an interface was described with
emphasis on how reflected and transmitted waves are generated. The behaviour at fractures
is further complicated since the fracture interface is not continuous as was the more general
case presented previously.
Instead the principle of diffraction becomes the most influential force. At the edges of fractures
the laws of reflection and refraction dont hold since there is a discontinuity. Instead the
incident seismic energy is radially scattered or diffracted [Kearey et al., 2002].
The reason for the diffractions can be explained by Huygens principle or the priciple of
a Fresnel zone [Sheriff, 1978]. Huygens principle states that every point on a wavefront
may be considered as a source of secondary wavefronts. A wave approaching the middle of
fracture would give rise to the reflected, refracated and transmitted waves. At the fracture
tips the wavefront bends around the fracture and according to Huygens principle new waves
are generated which propagate in all directions given rise to diffraction [Reynolds, 1997].
Diffraction therefor describes the transmission of energy by non-geometric ray paths.
Chapter 3
Numerical modelling
3-1
Numerical modelling is frequently used in geophysics. There are two types of modelling,
solving the forward problem, also called forward modelling, and solving the inverse problem,
known as inverse modelling. Forward modelling involves using a model where the parameters
are known or estimated to generate synthetic data. Inverse modelling describes the opposite
case where measured or observed data is inverted in order to predict the original model
parameters.
In the present study, where the initial fracture parameters will be defined and the effects on
seismic wave propagation are to be predicted, the forward problem is suitable. In order to turn
the initial parameters into a synthetic dataset, forward modelling applies a set of theoretical
equations related to the physical system being considered. These equations relate the model
parameters to the model data. For the case of seismic waves, the governing equation is the
wave equation presented in section 2-2. The input model parameters are the density and
velocity profiles representing the geological situation of interest.
Numerical forward modelling is of great importance for seismic exploration and interpretation. While the basic concept is clear, there are many different methods available for the
implementation of the forward modelling method. These include, among others, the finite
difference method, finite element method, pseudo spectral method, and the boundary element
method. The finite difference (FD) method is used extensively for the modelling of seismic
wave propagation [Aki and Larner, 1970] and [Virieux, 1986]. It has the advantages of being
a simple but powerful technique which is easy to implement in a modelling code.
Here the finite different method is also used, it involves using finite difference equations to
approximate the solutions of differential equations. More specifically for the seismic case,
the derivatives of the wave equation are replaced with finite differences based on the Taylor
expansion.
10
3-2
FDELMODC program
FDELMODC is an open source program which can model seismic waves conforming to the
2D wave equation. This approach is based on the work of [Robertsson et al., 1994] and
[Virieux, 1986]. To arrive to the wave equation for an acoustic medium, the first order linearized systems of Newton and Hookes law are considered, analogous to section 2-2 these are
the linearized equations of motion and deformation respectively.
Using a series of Taylor expansions, the approximations for the 1st order derivatives can
be obtained in both spatial coordinates and time. A 4th order approximation is used and
implemented in a staggered grid in the acoustic finite difference codes. In order to compute
the spatial derivatives of all grid points of a model grid, the combined eqns of Newtons and
Hookes law are progressively shifted through the model. At any one point, 4 grid points are
needed to calculate the partial derivative [Thorbecke, 2013].
Given this implementation of the Finite difference method, FDELMODC works by first reading the velocity and density information of a geological model in the form of a gridded input
model file. The lame parameters ( and ) are calculated using equations 3-1 and 3-2, where
k is the compressibility. Additionally cp is the P wave velocity, cs is the S wave velocity and
the density.
( + 2) = cp =
= cs
(3-1)
(3-2)
These values are used in the first order equations to calculate the wavefield at the next time
steps. Besides the lame parameters, information such as the grid spacing is required and this
will be explained along with the model set ups in the next section. However it is important
to mention that there are some restraints on the time step parameter depending on the grid
size chosen, due to the stability and dispersion criterion. These calculations are automatically
made so that only suitable values may be entered and used in the modelling experiment.
The program allows the implementation of 4 different finite difference schemes: acoustic,
visco acoustic, elastic and visco elastic. For each case there is a modification in the starting
equations of Newton and Hooke [Thorbecke, 2013]. However only the acoustic and elastic set
up will be used in this study. The program makes use of the Seismix Unix (SU) parameter
interface to read input files, to process and interpret the data and for making output files.
3-3
11
Modelling procedure
Following the explanation of how the finite difference program works, now the whole modelling
procedure will be described including all input parameters and considerations such as stability
and tapering in the FDELMODC stage.
There are 3 main stages which can be described, firstly a matlab script is used to set up the
fracture model, next FDELMODC is used to model the wave propagation, finally seismix unix
is used to display, characterise and interpret the data. The first 2 sections will be described
here following a brief introduction to the seismic survey design, the data interpretation will
follow in section 3-4.
3-3-1
The seismic survey is constructed to match the dimensions of the geological model, according
to the size of the grid in meters. Since the size of the model may change for different experiments there is no single standard survey design. However, the relative dimensions do remain
constant and shall be presented here.
The survey follows a conventional common mid point set up, with the source positioned at the
center of a receiver array at the surface. A second receiver array is placed at a depth of 4/5 of
the total depth of the model. The purpose of these arrays is to record the different response
of the fracture, i.e. the reflected and transmitted responses. The receiver spacing is 5m, and
this value remains the same for all grid sizes meaning that the resolution is constant. As a
result the total length of the arrays and hence the number of receivers does change depending
on the parameters chosen.
The source is defined as a ricker wavelet with a shift of 0.1 seconds, so as to avoid a source
signature at time t=0. This would introduce high frequencies and make the modelling dispersive. The source is defined differently depending on whether the acoustic or elastic scheme
is used. For the acoustic case a simple P wave is used with a pressure receiever. To be able
to examine S wave propagation, it was necessary to make use of the elastic scheme. In this
scheme the Fz source was chosen which emits both P and S waves, and behaves in a similar
manner to a vibroseis source. The elastic FD code takes into account all P and S wave conversions, and so as well as the P and S waves from the Fz source, Ps and Sp waves are also
generated.
3-3-2
A matlab script is used to design the fracture model. A grid of a certain size is defined for
example (1000x1000) pixels which will be the model space, and it is within this gird that the
fractures are placed in a defined way depending on the specification for the model.
All parameters related to the fractures and the surrounding rocks are defined at this stage.
The matlab script has the function of producing a density and a velocity matrix from the
starting model (see figure 3-1). These 2 output files will be used as the input data for the
FDELMODC stage of the modelling.
12
V (m/s)
Density (g/cm
200
400
200
200
400
400
600
600
800
800
600
800
1000
1200
1000
1400
1600
200
400
600
800 1000
1000
200
400
600
5000
800 1000
10000
1800
2000
200
400
(a)
600
800
1000 1200
x axis (grid points)
1400
1600
1800
Fracture model
2000
(b)
Velocity/Density profile
Figure 3-1: Examples of the matlab output files (a) Fracture model (b)Velocity/Density profile
3-3-3
The velocity and density models are acted upon by a UNIX scripts which calls FDELMODC
and preforms the finite difference modelling. A number of parameters related to the modelling
have to be defined within this script.
Firstly the grid spacing value in meters needs to be defined, this is in affect relating an actual
length to the grid size defined in the earlier matlab script. The numerical value of the grid
spacing multiplied by the size of the grid in pixels will determine the true size of the model
in meters. As such this needs to be considered while designing the matlab script.
Depending on the chosen grid spacing (dx/dz), the sampling time (t) needs to be defined
according to the criteria of equation 3-3, where vmax is the maximum velocity value:
t <
0.606(min(dx, dz))
vmax
(3-3)
This equation is related to the stability of the finite difference scheme [Thorbecke, 2013].
Other parameters such as number of sampling points (nt), the peak frequency and the modelling time (tmod) also need to be defined. The values for nt and tmod should be suitably
high depending on the sampling time and the size of the model. The peak frequency is chosen,
with the velocity and the the desired wavelength in mind.
Finally the model taper needs to be defined, in this case chosen to be an absorbing taper.
This is necessary in order to avoid reflections from the side of the model, which arise from
the artificial boundaries imposed on the model. In physical terms there would be no boundaries acting on the propagating waves, however since a numerical model has a limited size it
introduces an artifact to the data which needs to be removed. Different methods exist for
removing the edge affects, in this program a taper on the Vx and Vz fields is implmented
[Thorbecke, 2013].
13
The length of the taper is defined as a certain window of grid points. Since the value of
these grid points can be changed there is no set value of grid points which should be chosen.
However, the length of the taper in meters should be considered relative to the wavelength.
Generally a length of 2 times the wavelength is sufficient, however the longer the window the
longer the modelling will take.
A second unix script performs the same function as the first but for a homogeneous model, this
script is without fractures but using the same dimension and medium parameters. The result
of the homogeneous wave propagation is then subtracted from the result in the fractured
medium, in order to remove the direct wave travelling along the surface to leave only the
reflected and transmitted waves related with the fractures.
3-4
Data Analysis
The wavefield recorded at the recievers is exported as a data file, which can be processed with
Seismic UNIX (SU). Here, various seismic processing operations can be performed to generate
different plots so that the data may be interpreted in different ways. For the purposes of this
report, most of the interpretation is done in the space-time and space-frequency domains.
In the time domain, traces are selected out of the space-time receiever array to generate a
plot of seismic traces. A trace is generated for every receiver and records the response of the
Earth to the seismic wave propagating from the source. Any subsurface layer or diffractor
will effect the seismic wave on its path to the receiver.
In the frequency domain an amplitude spectra of frequencies is generated by performing a
fast Fourier transform on the space-time traces to produce complex frequency traces. An
additional operation is performed that takes the complex data and produces amplitude traces
as output. Finally the traces are stacked meaning that the frequncy spectras analysed in
this report are a superposition over all traces. This is done so that an average effect of the
fractures on the spectra is obtained.
14
Chapter 4
4-1
Preliminary modelling was previously carried out by a former student. The starting grid for
this was a 1000 by 1000 pixel grid with a 5m grid spacing, hence making a 5000m square
model. The issue with this set up is that the minimum possible fracture width is 5m, which
was deemed to be geologically unrealistic. In an attempt to model thinner fractures the grid
spacing value needs to made smaller while the total grid size is kept constant. There are two
main ways to accomplish this, the first is to change the dt, dz value in the modelling script,
alternatively the wavelength can be made relatively large so that the fractures are imaged
smaller.
Addressing the latter method first, the wavelength was previously set to be 250m. This value
is already large compared to common wavelengths used for exploration, considering a fracture
of 5m thick, the wavelength is 50 times the fracture width. It was decided to keep this value
for wavelength and focus on reducing the dt/dz. This value is easily changed in the modelling
script, however since this value is scaled with the numerical grid, the number of grid points
in the initial model needs to be increased proportionally so that the true grid size always
remains at 5000m. As the grid spacing value becomes slightly smaller the number of grid
points needed for modelling grows dramatically, resulting in a very large computation time
per model. After some test models were run, this approach was shown to be unfeasible for
efficient modelling since the run time for even a modestly small grid spacing took many hours.
15
16
4-2
A possible solution to this was to downscale the size of the total grid from 5000m to 1000m
resulting in smaller overall numerical grids and hence a shorter computation time. For the
time being the wavelength is kept the same as previously, however at 250m the wavelength
only fits into the 1000m model four times before it propagates through, which is not physically
realistic. Despite this, the modelling was attempted in order to examine the effects. A further
Amplitude
0.5
1.0
20
20
40
40
Frequency (Hz)
Frequency (Hz)
60
80
100
100
120
120
surface response
Amplitude
0.4
0.6
0.8
60
80
(a)
0.2
(b)
Figure 4-1: Homogeneous amplitude spectras for a 1000m grid and grid spacings of:
(red) 5m (green) 1m (blue) 0.5m (orange) 0.25m
problem was encountered relating to the taper size of the grid. The taper size was defined
for all models as 200 grid points. As with the model size, these points are scaled by the grid
spacing, so as this value decreases, the taper also gets smaller. For example a 1000m grid
with a grid spacing of 5m, results in a taper of 100m which is the same size as the model. If
however the grid spacing was reduced to 1m, the taper would only be 200m in length and 1/5
of the model. For all models except the 5m thick fracture the taper length is smaller than
the wavelength, this means the taper cannot accurately do its job of suppressing reflections
from the sides of the model.
The effects of this can be seen in the frequency amplitude spectras of Figure 4-1. These
figures are for the homogeneous case and as such the only difference between the different
curves displayed is the size of the grid spacing. In both figures it can be seen that the 5m grid
spacing, represented by the red curve, gives the smooth spectra that is normally obtained
from a ricker wavelet propagating through a homogeneous model. The same spectra should
be observed in all successively smaller grid spacing models, instead an attenuated behaviour
17
is observed, related to the taper not being sufficiently long enough to supress the model
boundary reflections.
Intuitively, there are two ways to overcome this problem, to make the taper bigger and the
wavelength smaller. Both of these methods may lead to more problems. Since decreasing the
wavelength will decrease the ratio between the wavelength and fracture width, the fractures
will be imaged larger, which was not the aim of the study. In order to support the wavelength
of 250m, the taper can be of course be increased, however for smaller grid spacings, a much
higher number of grid points is needed to maintain a taper of at least 2 times the wavelength.
This would again lead to the need for greater computation time, and the same time issues
are encountered as for the 5000m model.
4-3
Ringing effect
Considering the 2 models tested for the 250m wavelength, only the 5m thick fracture of both
models sizes was suitable for future modelling. Unfortunately when examining the modelling
results a further problem can be seen with the chosen grid parameters. In the snapshots of a
single 5m wide fracture of length 100m, a ringing affect can be observed inside the fracture
itself 4-2, tentatively attributed to waves propagating inside the fracture due to the ratio
between fracture size and wavelength.
X-axis (m)
1000
2000
3000
4000
1000
X Axis (m)
2000
3000
4000
1000
200
2000
400
Depth(m)
Depth (m)
3000
600
4000
800
(a)
Ringing effect
Ringing
1000
(b)
Zoomed in view
Figure 4-2: snapshot of wavefront and fracture interaction for a 5x500m fracture in a 5000m
square grid. (notice the ringing effect clearly observed inside the fracture
itself.)
18
To examine this in more detail, an experiment was made where the wavelength was kept
constant and only the velocity changed so that the impedance difference could be excluded as
a possible cause of the ringing. In order to do this, a smaller grid of 1000m with a 0.5m grid
spacing was looked at. Trying to balance the need for a suitably large taper length as well
as a reasonable computation time, the wavelength was selected to be 50m. Since this value is
100 times larger than the fracture width of 0.5 m, it is already an improvement on the 250m
wavlength with the 5m wide fracture, which was 50 times larger. Since the wavelength is
smaller, only an additional 400 grid points are used for the taper. Combining this in addition
to the grid points of the model, a total number of 2800 grid points is required in one dircetion
which is a reasonable amount to model.
Since the aim was to compare velocities relative to a fixed wavelength, 2 values were chosen
for the P wave velocity in the medium. The upper value was chosen to be 5000 m/s to give
a large contrast to the value inside the fractures which was 1650 m/s, Similarly 2000m/s
was chosen as the lower limit as these numbers are similar in value. In order to keep the
wavelength constant at 50m the frequency values needed to be changed and these work out to
be, 100hz and 40hz respectively for the 5000m/s and 2000m/s velocities. (Figure 4-3 displays
the snapshots for both examples. As with the original models and fracture width of 5m, the
5000m/s velocity model contained ringing which was not present in the same fracture imaged
with the same wavelength but with a lower P wave velocity of 2000m/s, as a result of this it
can be concluded that the ringing affect is a result of the velocity contrast rather than the
wavelength/fracture size ratio.
A further intermediate value of 3000m/s was chosen with the same frequency of 40hz as the
2000m/s velocity value. In order to achieve this the wavelength naturally increases slightly
to 75m. The results of these parameters were again checked for ringing and were found to be
clear of the affect. The motivation for doing this check was that now two different velocities
are made available, which are suitable for the modelling of both P and S waves.
4-4
Of course with greater computational resources, some of the problems described relating to
the CPU time may become obsolete, however from the results presented thus far there are
many limiting factors becoming apparent, particularly for a small scale study.
Conveniantly, the parameters which were found in the last section, are free of external affects such as ringing and provide a good balance between suitable modelling parameters and
practical issues such as time management. However if specific cases and values are needed
for modelling then more difficulties may arise. If it is indeed necessary to model with such a
case that contains ringing then another solution may be to manipulate the velocities inside
the fracture.
This method was examined to check whether it could be a viable solution. The velocity of
the material inside the fracture was made identical to the background velocity value and the
wavelength is not affected by the fracture. In the current model the velocity was changed
from 1690m/s to 5000m/s which was the background value. In order to compensate for this,
the value of the density needed to be increased from 1000 to 10,000 for inside the fracture.
With these values, the snapshot shown in figure 4-4 was shown to be free of ringing. It should
600
X-axis (m)
800
200
200
200
400
400
Depth (m)
Depth (m)
400
600
800
19
400
600
800
600
800
(a)
velocity 5000m/s
(b)
velocity 2000m/s
Figure 4-3: Wavefield propagation for the same fracture size, grid dimensions and wavelength
but varying background velocities. Notice the ringing effect which is still visible
in plot (a) for the 5000m/s velocity, is missing in plot (b) for the lower velocity
of 2000m/s. The diffraction event associated with the fracture boundary is still
observed although it is weaker in amplitude.
be noted, that since this solution was not going to be carried forward for future modelling,
it was not tested extensively for other density values and it is possible that new problems
may arise. It is offered here as a possible quick fix solution or a starting point for when the
parameters presented previously are not suitable for modelling.
20
X-axis (m)
1000
2000
3000
4000
Depth (m)
1000
2000
3000
4000
(a)
Figure 4-4: snapshot of wavefront and fracture interaction for a model where the velocity is
set to 0 inside the fracture. (Notice the ringing effect is now removed from
inside the fracture itself.)
Chapter 5
5-1
Source introduction
The next thing which was considered was the seismic source and the effect of fracture sensitivity to different waves, namely P and S waves. It is interesting to consider S waves in addition
to P waves as since they propagate by a different particle motion, it is possible that they may
provide extra information. This may be expecially true when a fracture were to contain a fluid
as S waves cannot propagate through fluids while P waves can. Here six different scenarios
are considered for each model, the first is a P wave response from a pure pressure source, the
second source is an Fz source which emits both P and S waves, both of these responses are
recorded at the surface as well as the P-s conversions which take place at a boundary. The
third example uses the same Fz source but with only the Ss wave contribution being recorded
at the surface. All 3 of these sets ups are run for a fracture that contains a solid material.
The 4th, 5th and 6th sets ups are for the same source receiver scenarios but for a liquid filled
fracture.
The modelling starts with a basic one fracture case, so as to get an idea of the effects of a
single fracture. The characteristics which will be examined are the orientation of the fracture,
the fracture length, and a varying fracture fill. Following this, the models will be built to be
more complex and geologically realistic.
Before, investigating any models with fractures included, the homogeneous case will be presented as it is to this that any changes are compared. Looking directly at the propagating
wavefield, the differences between the 2 sources are instantly noticable. The second case for
the Fz source, contains not only the P wave emitted from the first source but a second S wave
in addition.
21
22
The various waves are also shown clearly in the seismic traces. The reflected waves are
registered in the surface response, see figure 5-1. As is expected the first arrivals for the S
wave response arrive at later times than the first arrivals from the P source, since S waves
travel at a slower speed. For the Fz combined P and S response, the first arrivals are recorded
at later times than those for the pure P source, this is especially noticable at larger offsets
from the source. For the transmitted wave seen in the spectra at 800m, figure 5-2, the P wave
arrives at the same time in each case to the recievers. These first events can be seen clearly
to occur at arouns 0.3 seconds. For the Fz source examples, the S wave can then be seen to
arrive later than the P wave at around 0.4 seconds, since it travels at a slower speed.
The frequency response for both sources is similar since the same source signature is used.
In the amplitude spectras of figure 5-3, the same characteristics are observed for both source
types, in particular the shape of the spectra and peak frequency remain constant. The main
difference between the 3 is the change in amplitude which is observed. The transmitted
wavefield spectra is characterised by lower amplitudes than those at the surface, with the
peak amplitude for the transmitted P wave response about half of the value of the peak
amplitude of the reflected P wave.
0
20
40
trace number
60
80
100
20
40
trace number
60
80
100
0.2
0.2
0.2
40
trace number
60
80
100
Time (s)
0.1
Time (s)
0.1
Time (s)
0.1
20
0.3
0.3
0.3
0.4
0.4
0.4
(a)
(b)
Fz
(c)
Before examining the results of the fracture characteristics, the position of the fracture is
presented in order to give the results some context. The fracture used throughout this chapter
is kept constant in all other paramaters except that which is being tested in each case. The
fracture is positioned in the model at around 200m depth and at the 400m position in the
x-direction. Since the modelling grid is 1000m in length, the fracture is slightly offset from
the source towards the left hand side. For the orientation experiments the fracture is kept at
100m in length which is 2 times the wavelength. The tests for the different lengths are tested
for all three orientations, since 5 different lengths are tested there are 15 different runs for
each of the six source receiever scenarios described previously. The results for the orientation
experiments will be presented first followed by the length tests. When looking at the frequncy
amplitude spectra, keep in mind that it is the spectra averaged over the whole receiver array.
220
240
trace number
260
280
23
300
220
240
trace number
260
280
300
0.2
0.2
0.2
240
trace number
260
280
300
Time (s)
0.1
Time (s)
0.1
Time (s)
0.1
220
0.3
0.3
0.3
0.4
0.4
0.4
(a)
(b)
Fz
(c)
0.1
Amplitude
0.2
0.3
Amplitude
4
6
x10 -9
8
20
20
40
40
40
60
Frequency
20
Frequency
Frequency
60
0.5
Amplitude
1.0
x10 -9
1.5
60
80
80
80
100
100
100
120
120
120
(a)
(b)
Fz
(c)
5-2
5-2-1
Orientation effect
Frequency domain
Considering first the P reflection wave response at the surface(figure 5-4), the fracture has
the effect of lowering the observed amplitudes. The biggest deviation in spectra from the
homogeneous case is observed for the vertical fracture represented by the blue curve. For an
angle of this orientation, not only is the change in peak frequency the greatest but the spectra
also undergoes a significant change in shape, with a far broader spectra being displayed overall.
The spectra for a horizontal fracture is also far lower in amplitude than the homogeneous
case but the main characteristics are retained regarding the shape, and in particular the peak
frequency.
24
0.010
20
Frequency (Hz)
40
60
80
100
120
ture
0.5
20
20
40
40
60
x10 -9
Amplitude
1.5
Frequency (Hz)
Frequency (Hz)
x10 -10
Amplitude
1.0
0.5
1.0
60
80
80
100
100
120
120
ture
0.5
1.0
Amplitude
1.5
x10 -11
2.0
2.5
20
20
40
40
60
Frequency (Hz)
Frequency (Hz)
ture
0.5
Amplitude
1.0
1.5
x10 -10
2.0
60
80
80
100
100
120
120
ture
fracture
Turning to the Fz source with the complete seismic response being measured, the same
behaviour is observed regarding fracture orientation and the relative amplitudes, with again
the vertical fracture giving the smallest amplitudes. For this case the fracture oriented at 45
degrees and the horizontal fracture display a similar behaviour in terms of shape, and also in
amplitude value. The amplitude spectras as a whole are more complicated and are of much
smaller value to those for the P wave source which were on the order of 0.3.
0.005
25
Amplitude
0.010
0.015
0.020
20
Frequency (Hz)
40
60
80
100
120
ture
0.5
Amplitude
1.0
x10 -10
0
20
20
40
40
60
x10 -9
Amplitude
1.5
Frequency (Hz)
Frequency (Hz)
0.5
1.0
60
80
80
100
100
120
120
ture
0.5
Amplitude
1.0
1.5
x10 -11
2.0
20
20
40
40
60
Frequency (Hz)
Frequency (Hz)
ture
0.5
Amplitude
1.0
x10 -10
1.5
60
80
80
100
100
120
120
ture
fracture
Finally the Ss wave response, shows an even greater decrease in general amplitude values.
However it is interesting to note here that, the amplitude spectra for the horizontal fracture
is lowered by a greater amount relative to the previous cases as now the fracture inclined at
45 degrees displays the higher amplitude values. This is especially clear in the plot for a fluid
filled fracture, where the velocity inside the fracture was 0 for an S wave, since they cannot
travel through liquids. The same affect was also noticable although to a lesser degree in the
fluid filled fracture plot for the complete seismic response. This result can be attributed to
26
the particle motion by which the different waves propagate. Since the vertical fracture is in
the direction of wave propagation, when the P wave propagates by its longitudinal movement
with particle movement also in the direction of wave propagation, it is more sensitive to the
vertical fracture as the amplitudes in the frequency spectra are greatly lowered. When the S
wave propagates, with particle movement perpendicular to the wave propagation, the seismic
wave is now more sensitive to the horizontal fracture and this is seen in the results.
Figure 5-5, displays the same set up of figures but for the recievers at depth. For the P wave
transmission response, the spectras show much less change in shape compared to the homogeneous shape. In this case the blue curve representing the vertical fracture is not so damped
in value. While the horizontal and inclined angles keep their same relative relationship, the
amplitudes for the vertical fracture are much higher in comparison to the other two angles
and also to the vertical reflected response. The same thing is observed in the S wave reponse,
although it is not seen in the combined response.
In terms of the shape of the amplitude spectras, the transmitted spectras for the combined
response and also the s response are far more effected than for the P wave since they dont
retain the smooth homogeneous shape.
5-2-2
27
time domain
When considering the seismic traces for a fractured model, a difraction hyperbola is now
observed.
The seismic trace plots for P wave propagation and response (figure 5-6) is not so affected as
those for the combined P and S (figure 5-7) and S wave response (figure 5-9). This can be
observed in the figures where more events are observed for the latter two examples relating
to the different waves. In all cases the vertical fracture seems to give the greatest effect, since
two diffraction hyperbolaes are recorded, where the top and bottom of the 100m long fracture
are acting as point diffractors.
Considering first the P wave reponse (figure 5-6), following the first direct arrival, a second
higher amplitude second arrival can be observed. For the combined response (figures 5-7
and 5-8), the traces are much more complicated and more difficult to interpret easily. In the
horizontal fracture case, the 2 arrivals related to the arrival of the P and S waves are still
easily observed, where as for the other two angles there is a lot more interference. As was the
case for the P wave source, the vertical fracture results in many more events being recorded.
When considering the fluid fracture, there is minimal difference observed in the seismic traces.
For the S wave response (figure 5-9), the seismic section has similar characteristics to the
combined reponse. When comparing the solid and fluid filled fracture, it is seems that the
traces extend slightly longer in time for the fluid filled case. This effect is most noticable for
the vertical fracture since the events are higher in amplitude.
For the transmitted wavefields there is not much difference observed in the various plots
relating to different fracture angles for the P wave (see figure 5-11). The main thing to
mention is that for the horizontal fracture, the amplitude of the first arrivals is weaker for
later traces. However for the vertical fracture the first events are constant in amplitude
throughout the section, which is similar to the homogeneous case. This same effect is seen for
the Fz source as well. When comparing figure 5-14 to its homogeneous case, it can be seen
that the transmitted S wave is more attenuated than the P wave for receievers at this depth.
0
20
40
trace number
60
80
100
20
40
trace number
60
80
100
0.2
0.2
0.2
0.3
0.3
0.3
0.4
0.4
0.4
(a)
horizontal
40
trace number
60
80
100
Time (s)
0.1
Time (s)
0.1
Time (s)
0.1
20
(b)
45
(c)
vertical
28
20
40
trace number
60
80
100
20
40
trace number
60
80
100
0.2
0.2
0.2
0.3
0.3
0.3
0.4
0.4
0.4
(a)
40
trace number
60
80
100
Time (s)
0.1
Time (s)
0.1
Time (s)
0.1
20
horizontal
(b)
45
(c)
vertical
20
40
trace number
60
80
100
20
40
trace number
60
80
100
0.2
0.2
0.2
0.3
0.3
0.3
0.4
0.4
0.4
(a)
horizontal
40
trace number
60
80
100
Time (s)
0.1
Time (s)
0.1
Time (s)
0.1
20
(b)
45
(c)
vertical
5-3
Length effect
All three orientations are now further tested for a variety of lengths so as to examine this
effect. The lengths, chosen in relation to the P and S wavelengths are:
150m -2x S wavelength
100m -2x P wavelength
75m -S wavelength
50m -P wavelength
25m -1/2 or 1/3 wavelength
20
40
trace number
60
80
29
100
20
40
trace number
60
80
100
0.2
0.2
0.2
0.3
0.3
0.3
0.4
0.4
0.4
(a)
40
trace number
60
80
100
Time (s)
0.1
Time (s)
0.1
Time (s)
0.1
20
horizontal
(b)
45
(c)
vertical
20
40
trace number
60
80
100
20
40
trace number
60
80
100
0.2
0.2
0.2
0.3
0.3
0.3
0.4
0.4
0.4
(a)
horizontal
40
trace number
60
80
100
Time (s)
0.1
Time (s)
0.1
Time (s)
0.1
20
(b)
45
(c)
vertical
5-3-1
Frequency domain
Considering the P wave case first (figure 5-16), it can be seen that for the most simple
horizontal case, the amplitudes increase as fracture length increases and the characteristics
of the homogeneous case are maintained. For the fractured inclined at 45 degrees, the same
pattern is followed up untill the wavelength, but at lengths greater than the wavelength the
amplitudes start to get smaller.Also the frequency spectras for these larger fracture lengths
are more attenuated.
Consider the results for the vertical fracture, the behaviour here is different in that the
smallest length fracture shows the highest amplitude. However the spectras for all other
fracture lengths are extremely attenuated. This can be attributed to the fact the the ends of
the vertical fracture are acting as a point scatterer, this is dicussed further in the discussion
section (chapter 7).
30
220
240
trace number
260
280
300
220
240
trace number
260
280
300
0.2
0.2
0.2
0.3
0.3
0.3
0.4
0.4
0.4
(a)
240
trace number
260
280
300
Time (s)
0.1
Time (s)
0.1
Time (s)
0.1
220
horizontal
(b)
45
(c)
vertical
220
240
trace number
260
280
300
220
240
trace number
260
280
300
0.2
0.2
0.2
0.3
0.3
0.3
0.4
0.4
0.4
(a)
horizontal
240
trace number
260
280
300
Time (s)
0.1
Time (s)
0.1
Time (s)
0.1
220
(b)
45
(c)
vertical
For the combined response (figure 5-17) and the S wave response (figure 5-19), the trend
of increasing amplitude for increased fracture length is followed for both the horizontal and
inclined fracture, but the spectras are all generally a lot more attenuated than the P wave
case. For the vertical fracture, a similar behaviour is observed as for the vertical P wave
response.
When considering the transmitted wavefield, it can be observed that the amplitude spectras
are a lot less complicated in general. In the P wave case (figure 5-21), the curves for all lengths
are a lot less effected than the reflected wavefield. The curves retain their homogeneous shape,
and this is seen for all angles. The combined response (figure 5-22) and S wave response (figure
5-24), follow the same typical amplitude/fracture length trend.
220
240
trace number
260
280
31
300
220
240
trace number
260
280
300
0.2
0.2
0.2
0.3
0.3
0.3
0.4
0.4
0.4
(a)
240
trace number
260
280
300
Time (s)
0.1
Time (s)
0.1
Time (s)
0.1
220
horizontal
(b)
45
(c)
vertical
220
240
trace number
260
280
300
220
240
trace number
260
280
300
0.2
0.2
0.2
0.3
0.3
0.3
0.4
0.4
0.4
(a)
horizontal
240
trace number
260
280
300
Time (s)
0.1
Time (s)
0.1
Time (s)
0.1
220
(b)
45
(c)
vertical
5-3-2
time domain
Looking at the plots for the P wave source (figure 5-26) it is clear that the amplitudes of the
first arrivals decrease as the length increases. This effect is most noticable for a horizontal
fracture. When considering a vertical fracture, it was previously noted that two distinct
events could be noticed relating to the top and bottom of the fracture, from the fracture
length analysis it is shown that this occurs where the wavelength is greater than the fracture
length, as the first two plots corresponding to lengths of 25m and 50m do not display this
effect. Similarly for an inclined fracture at 45 degrees, for lengths equal or greater than the
wavelength an attenuation effect can be seen after the first arrival.
In terms of the transmitted wavefield for the P waves (figure 5-27), there is not much difference
observed between the different fracture lengths, except for a decrease in amplitudes and
greater attenuation as the fracture length increases. One thing that can be said however
32
220
240
trace number
260
280
300
220
240
trace number
260
280
300
0.2
0.2
0.2
0.3
0.3
0.3
0.4
0.4
0.4
(a)
240
trace number
260
280
300
Time (s)
0.1
Time (s)
0.1
Time (s)
0.1
220
horizontal
(b)
45
(c)
vertical
0.005
Amplitude
0.010
0.015
0.002
Amplitude
0.004
0.006
0.008
20
20
40
40
40
60
Frequency (Hz)
20
Frequency (Hz)
Frequency (Hz)
60
80
80
100
100
100
120
120
120
horizontal
Figure 5-16: P
source
(black)150m
(b)
Amplitude
0.002
0.003
0.004
0.005
60
80
(a)
0.001
45
(solid)(red)25m(green)50m
(c)
vertical
(blue)75m(orange)100m
is this effect seems to be more pronounced for the horizontal fracture than for the vertical
fracture, this was also true for the reflected wavefield.
For both the combined response (figure 5-28) and the S wave (figure 5-30) response, similar
effects regarding the decay in amplitudes can be observed. As mentioned before the vertical
fracture results in multiple events being recorded and these events stretch later in time as the
wavelength increases.
x10 -10
1.5
2.0
x10 -10
Amplitude
0
0.5
1.0
20
40
40
40
Frequency (Hz)
20
60
60
80
80
100
100
100
120
120
120
horizontal
(b)
60
80
(a)
x10 -11
Amplitude
1
20
Frequency (Hz)
Frequency (Hz)
Amplitude
1.0
0.5
33
45
(c)
vertical
Amplitude
0.5
1.0
x10 -9
1.5
x10 -9
Amplitude
0
0.5
1.0
20
20
40
40
40
60
Frequency (Hz)
20
Frequency (Hz)
Frequency (Hz)
60
80
80
100
100
100
120
120
120
horizontal
(b)
45
Amplitude
2
x10 -10
4
60
80
(a)
(c)
vertical
34
x10 -11
Amplitude
1
1.0
Amplitude
1.5
x10 -11
2.0
2.5
20
40
40
40
60
Frequency (Hz)
20
60
80
80
100
100
100
120
120
120
horizontal
Figure 5-19: S
response
(black)150m
Amplitude
0.5
(b)
45
(c)
(solid)(red)25m(green)50m
x10 -10
1.0
0.5
Amplitude
1.0
1.5
x10 -10
2.0
40
40
Frequency (Hz)
40
Frequency (Hz)
20
80
80
100
100
100
120
120
120
horizontal
Figure 5-20: S
response
(black)150m
(b)
x10 -11
6
60
80
(a)
vertical
Amplitude
20
60
(blue)75m(orange)100m
20
60
x10 -12
Amplitude
2
60
80
(a)
Frequency (Hz)
0.5
20
Frequency (Hz)
Frequency (Hz)
45
(fluid)(red)25m(green)50m
(c)
vertical
(blue)75m(orange)100m
0.005
Amplitude
0.010
0.015
0.002
Amplitude
0.004
0.006
0.008
0.010
20
20
40
40
40
60
Frequency (Hz)
20
60
80
80
100
100
100
120
120
120
horizontal
Figure 5-21: P
source
(black)150m
0.5
Amplitude
1.0
1.5
(b)
45
(c)
(solid)(red)25m(green)50m
x10 -10
2.0
0.2
0.4
Amplitude
0.6
1.0
40
40
40
Frequency (Hz)
20
Frequency (Hz)
20
60
60
80
80
100
100
100
120
120
120
(b)
45
0.025
vertical
Amplitude
4
x10 -11
6
60
80
horizontal
0.020
(blue)75m(orange)100m
x10 -10
0.8
20
(a)
Amplitude
0.010
0.015
0.005
60
80
(a)
Frequency (Hz)
Frequency (Hz)
Frequency (Hz)
35
(c)
vertical
36
0.5
Amplitude
1.0
x10 -9
1.5
x10 -9
Amplitude
0
0.5
1.0
20
20
40
40
40
60
Frequency (Hz)
20
Frequency (Hz)
Frequency (Hz)
60
80
80
100
100
100
120
120
120
horizontal
(b)
0.8
x10 -9
1.0
60
80
(a)
Amplitude
0.4
0.6
0.2
45
(c)
vertical
0.5
Amplitude
1.0
x10 -11
1.5
Amplitude
0
0.5
1.0
x10 -11
1.5
Amplitude
20
40
40
40
Frequency (Hz)
20
60
60
80
80
100
100
100
120
120
120
horizontal
Figure 5-24: S
response
(black)150m
(b)
x10 -11
3
60
80
(a)
20
Frequency (Hz)
Frequency (Hz)
45
(solid)(red)25m(green)50m
(c)
vertical
(blue)75m(orange)100m
x10 -10
Amplitude
0.5
1.0
x10 -10
Amplitude
0
0.5
1.0
20
20
40
40
40
60
Frequency (Hz)
20
Frequency (Hz)
Frequency (Hz)
37
60
80
80
100
100
100
120
120
120
horizontal
Figure 5-25: S
response
(black)150m
(b)
Amplitude
1.0
x10 -10
1.5
60
80
(a)
0.5
45
(fluid)(red)25m(green)50m
(c)
vertical
(blue)75m(orange)100m
38
20
40
trace number
60
80
100
20
40
trace number
60
80
100
20
40
trace number
60
80
100
20
40
trace number
60
80
100
0.2
0.2
0.2
0.2
0.2
0.3
0.3
0.3
0.3
0.3
0.4
0.4
0.4
0.4
0.4
40
trace number
60
80
100
Time (s)
0.1
Time (s)
0.1
Time (s)
0.1
Time (s)
0.1
Time (s)
0.1
20
horizontal (b)
horizontal (c)
horizontal (d)
horizontal (e)
horizontal
length=25m
length=50m
length=75m
length=100m
length=150m
(a)
20
40
trace number
60
80
100
20
40
trace number
60
80
100
20
40
trace number
60
80
100
20
40
trace number
60
80
100
0.2
0.2
0.2
0.2
0.2
0.3
0.3
0.3
0.3
0.3
0.4
0.4
0.4
0.4
0.4
40
trace number
60
80
100
Time (s)
0.1
Time (s)
0.1
Time (s)
0.1
Time (s)
0.1
Time (s)
0.1
20
inclined (g)
inclined (h)
inclined (i)
inclined (j)
inclined
length=25m
length=50m
length=75m
length=100m
length=150m
(f)
20
40
trace number
60
80
100
20
40
trace number
60
80
100
20
40
trace number
60
80
100
20
40
trace number
60
80
100
0.2
0.2
0.2
0.2
0.2
0.3
0.3
0.3
0.3
0.3
0.4
0.4
0.4
0.4
0.4
40
trace number
60
80
100
Time (s)
0.1
Time (s)
0.1
Time (s)
0.1
Time (s)
0.1
Time (s)
0.1
20
vertical (l)
vertical (m)
vertical (n)
vertical (o)
vertical
length=25m
length=50m
length=75m
length=100m
length=150m
(k)
220
240
trace number
260
280
300
220
39
240
trace number
260
280
300
220
240
trace number
260
280
300
220
240
trace number
260
280
300
0.2
0.2
0.2
0.2
0.2
0.3
0.3
0.3
0.3
0.3
0.4
0.4
0.4
0.4
0.4
240
trace number
260
280
300
Time (s)
0.1
Time (s)
0.1
Time (s)
0.1
Time (s)
0.1
Time (s)
0.1
220
horizontal (b)
horizontal (c)
horizontal (d)
horizontal (e)
horizontal
length=25m
length=50m
length=75m
length=100m
length=150m
(a)
220
240
trace number
260
280
300
220
240
trace number
260
280
300
220
240
trace number
260
280
300
220
240
trace number
260
280
300
0.2
0.2
0.2
0.2
0.2
0.3
0.3
0.3
0.3
0.3
0.4
0.4
0.4
0.4
0.4
240
trace number
260
280
300
Time (s)
0.1
Time (s)
0.1
Time (s)
0.1
Time (s)
0.1
Time (s)
0.1
220
inclined (g)
inclined (h)
inclined (i)
inclined (j)
inclined
length=25m
length=50m
length=75m
length=100m
length=150m
(f)
220
240
trace number
260
280
300
220
240
trace number
260
280
300
220
240
trace number
260
280
300
220
240
trace number
260
280
300
0.2
0.2
0.2
0.2
0.2
0.3
0.3
0.3
0.3
0.3
0.4
0.4
0.4
0.4
0.4
240
trace number
260
280
300
Time (s)
0.1
Time (s)
0.1
Time (s)
0.1
Time (s)
0.1
Time (s)
0.1
220
vertical (l)
vertical (m)
vertical (n)
vertical (o)
vertical
length=25m
length=50m
length=75m
length=100m
length=150m
(k)
40
20
40
trace number
60
80
100
20
40
trace number
60
80
100
20
40
trace number
60
80
100
20
40
trace number
60
80
100
0.2
0.2
0.2
0.2
0.2
0.3
0.3
0.3
0.3
0.3
0.4
0.4
0.4
0.4
0.4
40
trace number
60
80
100
Time (s)
0.1
Time (s)
0.1
Time (s)
0.1
Time (s)
0.1
Time (s)
0.1
20
horizontal (b)
horizontal (c)
horizontal (d)
horizontal (e)
horizontal
length=25m
length=50m
length=75m
length=100m
length=150m
(a)
20
40
trace number
60
80
100
20
40
trace number
60
80
100
20
40
trace number
60
80
100
20
40
trace number
60
80
100
0.2
0.2
0.2
0.2
0.2
0.3
0.3
0.3
0.3
0.3
0.4
0.4
0.4
0.4
0.4
40
trace number
60
80
100
Time (s)
0.1
Time (s)
0.1
Time (s)
0.1
Time (s)
0.1
Time (s)
0.1
20
inclined (g)
inclined (h)
inclined (i)
inclined (j)
inclined
length=25m
length=50m
length=75m
length=100m
length=150m
(f)
20
40
trace number
60
80
100
20
40
trace number
60
80
100
20
40
trace number
60
80
100
20
40
trace number
60
80
100
0.2
0.2
0.2
0.2
0.2
0.3
0.3
0.3
0.3
0.3
0.4
0.4
0.4
0.4
0.4
40
trace number
60
80
100
Time (s)
0.1
Time (s)
0.1
Time (s)
0.1
Time (s)
0.1
Time (s)
0.1
20
vertical (l)
vertical (m)
vertical (n)
vertical (o)
vertical
length=25m
length=50m
length=75m
length=100m
length=150m
(k)
220
240
trace number
260
280
300
220
41
240
trace number
260
280
300
220
240
trace number
260
280
300
220
240
trace number
260
280
300
0.2
0.2
0.2
0.2
0.2
0.3
0.3
0.3
0.3
0.3
0.4
0.4
0.4
0.4
0.4
240
trace number
260
280
300
Time (s)
0.1
Time (s)
0.1
Time (s)
0.1
Time (s)
0.1
Time (s)
0.1
220
horizontal (b)
horizontal (c)
horizontal (d)
horizontal (e)
horizontal
length=25m
length=50m
length=75m
length=100m
length=150m
(a)
220
240
trace number
260
280
300
220
240
trace number
260
280
300
220
240
trace number
260
280
300
220
240
trace number
260
280
300
0.2
0.2
0.2
0.2
0.2
0.3
0.3
0.3
0.3
0.3
0.4
0.4
0.4
0.4
0.4
240
trace number
260
280
300
Time (s)
0.1
Time (s)
0.1
Time (s)
0.1
Time (s)
0.1
Time (s)
0.1
220
inclined (g)
inclined (h)
inclined (i)
inclined (j)
inclined
length=25m
length=50m
length=75m
length=100m
length=150m
(f)
220
240
trace number
260
280
300
220
240
trace number
260
280
300
220
240
trace number
260
280
300
220
240
trace number
260
280
300
0.2
0.2
0.2
0.2
0.2
0.3
0.3
0.3
0.3
0.3
0.4
0.4
0.4
0.4
0.4
240
trace number
260
280
300
Time (s)
0.1
Time (s)
0.1
Time (s)
0.1
Time (s)
0.1
Time (s)
0.1
220
vertical (l)
vertical (m)
vertical (n)
vertical (o)
vertical
length=25m
length=50m
length=75m
length=100m
length=150m
(k)
42
20
40
trace number
60
80
100
20
40
trace number
60
80
100
20
40
trace number
60
80
100
20
40
trace number
60
80
100
0.2
0.2
0.2
0.2
0.2
40
trace number
60
80
100
Time (s)
0.1
Time (s)
0.1
Time (s)
0.1
Time (s)
0.1
Time (s)
0.1
20
0.3
0.3
0.3
0.3
0.3
0.4
0.4
0.4
0.4
0.4
horizontal (b)
horizontal (c)
horizontal (d)
horizontal (e)
horizontal
length=25m
length=50m
length=75m
length=100m
length=150m
(a)
20
40
trace number
60
80
100
20
40
trace number
60
80
100
20
40
trace number
60
80
100
20
40
trace number
60
80
100
0.2
0.2
0.2
0.2
0.2
0.3
0.3
0.3
0.3
0.3
0.4
0.4
0.4
0.4
0.4
(f)
length=25m
0
20
40
80
100
20
40
trace number
60
80
100
20
40
trace number
60
80
100
20
40
trace number
60
80
100
80
100
0.2
0.2
0.2
0.2
0.2
20
40
trace number
60
80
100
Time (s)
0.1
Time (s)
0.1
Time (s)
0.1
Time (s)
0.1
Time (s)
0.1
0.3
0.3
0.3
0.3
0.3
0.4
0.4
0.4
0.4
0.4
length=25m
trace number
60
inclined (g)
inclined (h)
inclined (i)
inclined (j)
inclined
length=50m
length=75m
length=100m
length=150m
trace number
60
(k)
40
Time (s)
0.1
Time (s)
0.1
Time (s)
0.1
Time (s)
0.1
Time (s)
0.1
20
vertical (l)
vertical (m)
vertical (n)
vertical (o)
vertical
length=50m
length=75m
length=100m
length=150m
220
240
trace number
260
280
300
220
240
trace number
260
280
300
220
240
trace number
260
280
43
300
220
240
trace number
260
280
300
0.2
0.2
0.2
0.2
0.2
240
trace number
260
280
300
Time (s)
0.1
Time (s)
0.1
Time (s)
0.1
Time (s)
0.1
Time (s)
0.1
220
0.3
0.3
0.3
0.3
0.3
0.4
0.4
0.4
0.4
0.4
horizontal (b)
horizontal (c)
horizontal (d)
horizontal (e)
horizontal
length=25m
length=50m
length=75m
length=100m
length=150m
(a)
220
240
trace number
260
280
300
220
240
trace number
260
280
300
220
240
trace number
260
280
300
220
240
trace number
260
280
300
0.2
0.2
0.2
0.2
0.2
0.3
0.3
0.3
0.3
0.3
0.4
0.4
0.4
0.4
0.4
(f)
length=25m
0
220
240
280
300
220
240
trace number
260
280
300
220
240
trace number
260
280
300
220
240
trace number
260
280
300
280
0.2
0.2
0.2
0.2
0.2
0.3
0.3
0.3
0.3
0.3
0.4
0.4
0.4
0.4
0.4
220
240
trace number
260
280
300
vertical (l)
vertical (m)
vertical (n)
vertical (o)
vertical
length=50m
length=75m
length=100m
length=150m
5-4
300
Time (s)
0.1
Time (s)
0.1
Time (s)
0.1
Time (s)
0.1
Time (s)
0.1
length=25m
trace number
260
inclined (g)
inclined (h)
inclined (i)
inclined (j)
inclined
length=50m
length=75m
length=100m
length=150m
trace number
260
(k)
240
Time (s)
0.1
Time (s)
0.1
Time (s)
0.1
Time (s)
0.1
Time (s)
0.1
220
Referring back to the previous plots for comparing orientations and lengths of the fracture it
is possible to make some observations on the difference between the P and S propagation and
the fracture fills. First the fracture fill is described. Looking for example at figure 5-4, In the
frequency domain it is possible to notice a clear shift in peak frequencies to lower frequencies
when the fracture contains fluid. Also there is far less variation in the lower frequencies when
comparing to the solid case. Also for all angles there is a strong decrease in general amplitudes
when switching from the solid fill to the liquid fill. In the time domain, it is much harder
to notice any significant effect caused by the fracture fill. When comparing between P and
S waves, in the frequency domain it can be noticed that the P wave gives a much smoother
44
amplitude spectra where as the S wave spectra tends to be more attenuated. There is also a
difference is sensitivity between the two waves for different orientations of fracture.
Chapter 6
6-1
6-1-1
parallel fractures
model definition
It is well known that fracture clusters often occur as sets of parallel joints through the rock
mass [Hsu and Schoenberg, 1993]. [Schoenberg and Douma, 1988], stated that theoretically,
the effect of parallel fractures which are large compared to the wavelength, is similar to a
small dilute concentration of aligned microcracks.
The first complex model which was modelled was for parallel sets of fractures with different
spacings between the fractures. First, the amount of fractures was set at 2 and then increased
to 5, 10, 20, 30, 50, 70 and 100 fractures. Each of these data sets were tested for 5 different
normal spacing values, 3m, 5m, 10m, 30m and 50m. Figure 6-1, show 2 examples of how the
model is constructed.
6-1-2
Considering the amplitude spectras, the first noticable feature is that for lower densities of
fractures (< 30 fractures)(see figure 6-2(a) - 6-2(e)), the amplitude spectras are markedly
more different for different spacings between the fractures, especially with regards to the
shape 6-2.
For the higher fracture amounts (> 50 fractures) the spectras are much more similar with
only slight differences noticed with regards to peak frequencies.
45
200
300
300
400
400
46
500
500
600
600
700
700
800
200
400
600
800
1000 1200
x axis (grid points)
1400
1600
1800
800
2000
200
400
600
800
1000 1200
x axis (grid points)
1400
1600
1800
2000
0.002
0.004
Amplitude
0.006 0.008
Amplitude
0.010
0.012
0.01
0.02
0.01
Amplitude
0.02
0.03
0.04
20
40
40
40
40
60
60
Frequency (Hz)
20
Frequency (Hz)
20
60
80
80
80
80
100
100
100
120
120
120
120
0.01
Amplitude
0.02
(b) 5 fractures
0.03
0.01
Amplitude
0.02
0.03
0.01
Amplitude
0.02
0.03
40
60
Frequency (Hz)
20
40
Frequency (Hz)
20
40
Frequency (Hz)
20
40
60
0.03
0.01
Amplitude
0.02
0.03
0.04
60
80
80
80
80
100
100
100
100
120
120
120
120
(e) 30 fractures
Amplitude
0.02
20
60
0.01
60
100
(a) 2 fractures
Frequency (Hz)
20
Frequency (Hz)
Frequency (Hz)
For plots 6-2(a) - 6-2(e), it can be seen that the red, green and blue curves (3,5 and 10
m spacing) follow the same trend of decreasing amplitudes with increasing spacing. When
the spacing reaches 30m the behaviour changes with a shift in peak amplitude to higher
Amplitude
1.0
1.5
2.0
x10 -3
2.5
0.002
0.006
0.005
Amplitude
0.010
0.015
0.020
0.025
20
40
40
40
40
60
60
Frequency (Hz)
20
Frequency (Hz)
20
60
80
80
80
80
100
100
100
120
120
120
120
(b) 5 fractures
Amplitude
0
0.02
0.04
0.02
Amplitude
0.04
0.06
Amplitude
0.08
0.05
0.10
0.15
40
60
Frequency (Hz)
20
40
Frequency (Hz)
20
40
Frequency (Hz)
20
40
60
Amplitude
0.02
0.03
0.04
0.05
20
60
0.01
60
100
(a) 2 fractures
Frequency (Hz)
Amplitude
0.004
20
Frequency (Hz)
Frequency (Hz)
47
0.05
Amplitude
0.10
0.15
60
80
80
80
80
100
100
100
100
120
120
120
120
frequencies as observed in the orange curve, for even greater spacing of 50m the peak frequency
becomes more consistent with the general trend. This behaviour is consistent for all values
up to 30 fractures.
When considering the transmitted wavefield (figure 6-3), the amplitudes displayed are now
much higher, and for the cases of > 20 fractures the curves now display a smoother homogeneous shape. The trend that was display for the reflected wavefield of decreasing amplitude
with increasing spacing, is present only for the amounts of fractures less than 20. For the
amounts of fractures greater than 30 the trend is now reversed with increasing amplitude for
increasing spacing.
Note that as the number of fractures increases some of the curves are missing from the
spectras, this is because they represent higher spacings between fractures and as such for
greater amounts of fractures, the biggest spacings dont fit into the model.
48
6-2
Different distributions
6-2-1
model definition
[Vlastos et al., 2003], already showed that fracture clustering can have a significant effect on
the wavefield energy attenuation and that different distributions of fractures have a significant
influence on multiple scattering. It has also been acknowledged that a medium becomes
azimuthally anisotropic with respect to wave propagation, when fracture networks have a
preferred orientatoin [Bakulin et al., 2000]. In the [Vlastos et al., 2003] paper, the effects
of fracture position within the medium were examined in more detail while the fractures
themselves were randomly oriented.
The purpose of this model was to examine the effect of switching from a case where fractures
were randomly distributed to when fractures where set into fixed regular positions. Aside from
this we were also interested in how the wavefield behaves when a large amount of fractures
are randomly oriented or orineted in the same preferential direction. To achieve this, 300
fractures were considered, for the following 3 distributions:
1) random angles and random distribution
2) constant angle and random distributions
3) Fixed positions
400
600
600
800
1000
1200
400
600
Depth (grid points)
200
400
200
200
800
1000
1200
1400
1400
1600
1600
1800
1800
800
1000
1200
2000
200
400
600
800
1000 1200
x axis (grid points)
1400
(a) random
6-2-2
1600
1800
2000
2000
1400
200
400
600
800
1000 1200
x axis (grid points)
1400
1600
1800
2000
200
400
600
800
1000
x axis (grid points)
1200
1400
(c) regimented
The seismic traces in the time domain show some differences between the randomly oriented
fractures and those with a fixed constant orientation. In the case of randomly oriented
fractures, the first event has a lower amplitude than in the case where the fractures all have
the same orientation. Also for the random case, there is a lot more multiple events following
the first event, these may be interpreted as coda waves, see figure 6-4. The response for
the transmitted field was so similar to what was described for the reflected field, that it was
decided not to display the seismic traces for the receivers at depth.
6-2-3
In the frequency domain, the incremental change in angle and then also fracture position
seemed to have a minimal affect. This may possibly be due to the large number of fractures
20
40
trace number
60
80
100
20
49
40
trace number
60
80
100
0.2
0.2
0.2
0.3
0.3
0.3
0.4
0.4
0.4
(d) random
40
trace number
60
80
100
Time (s)
0.1
Time (s)
0.1
Time (s)
0.1
20
(f) regimented
Figure 6-4: Surface response (a) random (b) random distributions (c)Fixed positions and
orientations
considered, since the interference between fractures may mask most individual effects.
The frecuency plot of figure 6-5 shows very little change in shape for the amplitude spectra,
and it can be seen that the amplitude increases only slightly as the fractures change from
randomly oriented to a fixed angle of 45 degrees.
The model for the fractures with fixed positions and orientations can also be seen as an
extension of the model presented in section 6-1, since that was for a layer of parallel fractures
and here the model is filled with multiple layers of parallel fractures. Comparing the curve
for the fixed fractures with the earlier one for a layer of parallel fractures, there is a very large
increase in amplitude now that the whole model is filled with fractures.
The transmitted wavefield is also displayed in figure 6-5. The frequency spectra here at depth
is much less effected for all three models, with higher amplitudes than the reflected case and
a smoother shape. Opposite to the reflected case, the amplitude increases as the fracture
distribution becomes more random.
50
0.05
Amplitude
0.15
0.20
20
20
40
40
60
80
100
100
120
120
Reflected wavefield
0.2
0.3
60
80
(a)
0.1
Frequency (Hz)
Frequency (Hz)
Amplitude
0.10
(b)
Transmitted wavefield
Figure 6-5: (red) random (green) random distributions (blue)Fixed positions and orientations
6-3
6-3-1
preferential axis
model definition
The intersection of fractures is a common geological behaviour and the angle of intersection
tends to be a predictable behaviour. Small scale shear fractures for example nearly always
intersect at an angle of around 60 degrees [Twiss and Moores, 1992]. Due to the stress field
acting on a rock, fractures tend to occur in a preferred direction, this direction is the plane
where the shear stress is maximum [Ranalli, 1995]. It is not uncommon for a rock to have two
different preferred axis of fracture, however these are normally found roughly perpendicular
to each other (or for example at an angle of 60 degrees).
Here a model is created with preferred orientation along two axis at 90 degrees to each other.
There are 3 variations of the model. The first has 70 percent of the fractures oriented at 45
degrees and the remaining fractures at an angle perpendicular to this. The second model is
with half of the fractures on each axis and the final model is for the opposite case of the
first. The model for the 50/50 percentage split is shown in figure ??. A summary of the
distributions is given:
1) 70/30 *percentage split
2) 50/50 *percentage split
51
200
400
600
800
1000
1200
1400
1600
1800
2000
200
400
(a)
6-3-2
600
800
1000 1200
x axis (grid points)
1400
1600
1800
2000
Figure 6-6 displays the seismic traces for the surface response. The third plot for the 70/30
percentage split, displays some distortion compared to the other two plots. Besides this there
is not much information gained from these results in the time domain.
As was the case in section 6-2-2, the transmitted wavefield does not give any additional
features of interest.
0
20
40
trace number
60
80
100
20
40
trace number
60
80
100
0.2
0.2
0.2
0.3
0.3
0.3
0.4
0.4
0.4
(b)
30/70
40
trace number
60
80
100
Time (s)
0.1
Time (s)
0.1
Time (s)
0.1
20
(c)
50/50
(d)
70/30
Figure 6-6: Surface response (a) 30/70 (b) 50/50 (c) 70/30
52
6-3-3
When looking at the comparison plot 6-7, the 3 models have similar amplitude spectras,
especially for the lower fequencies.
An interesting feature is that the curves for the 50/50 split and the 70/30 split have very
similar amplitudes and this is also true in the transmitted response.
Where as the curve for the 30/70 split has a loss in high frequncies for the reflected response
and also a much lower peak frequency. In the transmitted response it has a higher amplitude
than the other 2 distributions. This is most propabably related to the distortion observed in
the seismic traces of figure 6-6.
Another interesting thing to note is the low amplitudes as a whole when condsidering this
model of fractures along 2 axis. Previously it was shown that for the same number of fractures
randomly oriented and fixed at 45 degrees, the amplitudes where not only much higher in
value but also similar in value to each other. It might be expected that fractures along 2
orientations namely 45 degrees and 135 degrees, would also give a similar response which is
not the case. However, these vaules are still larger than those seen for the single layer of
parallel fractures oriented at 45 degrees.
0.02
0.08
20
20
40
40
60
80
100
100
120
120
Reflected response
0.10
60
80
(a)
Amplitude
0.05
Frequency (Hz)
Frequency (Hz)
Amplitude
0.04
0.06
(b)
Transmitted response
Chapter 7
Discussion of results
Finite difference modelling has been used to investigate the effects of fractures on seismic
wave propagation. Features of interest were in particular fracture length, orientation and distribution. For P wave propagation, these characteristics have already been largely examined
in the literature, hence in this study S wave propagation was also incorporated since it was
expected to give extra information compared to the P waves analysis, especially in the time
domain. In order to make a relaible comparison between the different waves it was necessary
to look at the effects on a single fracture, this was especially true since a new starting model
was used compared to the previous work.
Many authors have previously examined the effect of fractures on the velocity and attenuation
of seismic waves including [Boadu and Long, 1997] among others.In this study it was shown
that the presence of fractures in a medium lowers the amplitudes in the frequency domain,
regardless of the fracture characteristics or the wave type being considered. For the reflected
wavefield, a vertical fracture always had the most pronounced effect on the amplitude spectra,
causing a dramatic decrease in amplitude and a strong attenuation, and this was true for
both wave types. However, for the transmitted wavefield it was shown that the vertical
fracture showed the least change, as it displayed the highest amplitudes out of the 3 different
orientations.
Condsidering the length effect in the frequency domain, there seemed to be a correlation between increasing length and increasing amplitude, this was most noticable for the horizontal
fracture and also for the transmitted wavefield. However, for a vertical fracture, this relationship was a bit more complicated since at lengths equal or greater to the wavelength, there
was a very strong attenuation and damping of the amplitude values. As the fracture gets
longer, greater attenuation takes place. For the orientation and length effects in the frequncy
domain little difference was shown between p and s waves. Therefore in order to display the
effects, Figure 7-1 relates to the P wave amplitude traces, since they show the effect in a more
clear manner.
Although the S wave displayed similar behaviours to what was described above, there were
some subtle differences worth noting. In the P wave case the angle oriented at 45 degrees
also displayed attenuation for fracture lengths greater than the wavelength, however this was
not the case for the s wave which followed the typical length/amplitude trend for all fracture
lengths. Also for the s wave reflected frequency response, there was a shift in peak frequency
53
54
to lower frequencies for the curve relating to the fracture length equal to the wavelength of
the S wave, but for all other fracture lengths, greater or smaller than this the original peak
frequency was maintained. For the transmitted case the effect was opposite with the fracture
length equal to the wavlength keeping its original peak frequency and all other curves shifting
to lower frequencies, figure 7-2.
0.005
Amplitude
0.010
0.015
20
20
40
40
Frequency (Hz)
Frequency (Hz)
60
0.001
Amplitude
0.002
0.003
0.004
0.005
60
80
80
100
100
120
120
reflected response:
(b) reflected response:
horizontal
vertical
(a)
0.005
Amplitude
0.010
0.015
20
20
40
40
60
Frequency (Hz)
Frequency (Hz)
0.005
Amplitude
0.010
0.015
0.020
60
80
80
100
100
120
120
transmitted
re(d) transmitted
sponse: horizontal sponse: vertical
(c)
0.025
re-
The relationship between increasing fracture length and increasing amplitude can be explained
by the fact that a greater amount of energy is reflected back to the surface and that the fracture
is sensed more as a seperate interface than a diffractor for longer lengths.
For the smallest length of 25m (1/2 wavelength) there is less reflection and more scattering
and diffractions from the fracture resluting in the lowering of amplitude.
In agreement with [Vlastos et al., 2003], the longer fractures equal or greater to the wavelength result in a more complicated wavefield than that produced by a shorter fracture since
they as seen by the propagating wavefield as an interface or reflector. Smaller fractures are
55
x10 -11
Amplitude
1
20
20
40
40
60
Frequency (Hz)
Frequency (Hz)
0.5
Amplitude
1.0
x10 -11
1.5
60
80
80
100
100
120
120
reflected response:
(b) transmitted
response:
horizontal
horizontal
(a)
Figure 7-2: S
response
(black)150m
(solid)(red)25m(green)50m
(blue)75m(orange)100m
recognised more as point scatterers, and generate secondary wavefields, in this case there is
a clear image of P and S wave propagatio through the medium.
Considering the seismice traces for both P and S waves a greater effect is shown for the reflected wavefield than for the transmitted. An immediate observation is that vertical fractures
give a strong effect in the time domain when they are equal or greater in length than the
wavelength. The result is very noticable since 2 diffraction hyperbolae are produced.This is
caused by the ends of the fracture acting as separate point diffractors. Since the fracture
length is relatively long (100m) compared to the wavelength, the effect is very obvious. The
vertical fracture is also characterised by constant high amplitudes across the hyperbola. This
is true of both P and S waves, however, the traces for the s wave show more events than the
P wave traces after the first arrival. However this effect is not so noticable for the recievers
at a depth of 800m.
With regards to fracture length it was shown that as the length increases the amplitudes of
the events become weaker. In the case of the vertical fracture, the distance between the two
hyperbolae increases with the length, for a fracture length of 25m (1/2 wavelength), only 1
hyperbola can be seen.
From seismic wave theory, there are some observations which we might intuitively expect to
see. The first is related to the compressional movement of p waves and the shear motion of
s waves. The individual particle that make up the wave propagate along the direction of the
wave for p waves and at right angles to the wave motion for s waves. As a result it should
be possible to notice weaker amplitudes of the p event for situations where the fracture is
in the direction of the wave propagation, and weaker S wave events where the fracture is
perpendicular to the wave direction [Vlastos et al., 2007].
However figure 7-3 displays the traces for the Fz combined response for the horizontal and
vertical case and this effect is not clearly visible. In fact the opposite seems to be true with
the p event being stronger in the horizontal fracture and the s event more dominant for the
56
vertical fracture.
The second event which was of interest is the fact that s waves are predicted to show significant time delays in the onset of arrivals due to a fractured medium [King and Xu, 1989]
[Hardage, 2011]. Unfortunately in the results presented in this study, this was not seen clearly
in the time domain results either for the angle or the orientation.
0
220
240
trace number
260
280
300
0.2
0.2
240
trace number
260
280
300
Time (s)
0.1
Time (s)
0.1
220
0.3
0.3
0.4
0.4
(a)
horizontal
(b)
Analysis in the time domain corroborates the findings about a single fracture. This is especially clear for the vertical fracture which produces two separate diffraction hyperbolae.
This is due to the fracture being seen as an interface where the two end points give seperate
responses. For the fracture smaller than the wavelength only one hyperbola is seen since the
fracture is acting as a point scatterer.
Following the P and S wave analysis, more complex models were built in order to examine
the effects of wave propagation.
First the fracture fill was changed to approximate a solid filled fracture and a fracture which
contained a fluid. The main noticable difference was that for the s wave frequency reponse,
there was a clear shift to lower peak frequencies and a lowering of spectra amplitudes for the
fluid case, which was independent of angle.
The next models which were considered were sets of parallel 2D fractures with a varying
normal spacing being implemented between the individual fractures. In this way a fractured
layer is produced. As the number of fractures is increased there is a slight increase in overall
amplitude, although this increase is very minimal.
For sets of smaller amounts of fractures (<30 fractures), the amplitudes were greater attenuated as the spacing between the fractures was increased. However once the spacing between
the fractures became too large (approximately 30m) the attenuation became less and there
was a shift in peak frequency to higher values.
For even greater amounts of fractures (>30), there was far less variation in the frequency
spectras when considering different spacings. The spectras are all heavily attenuated, and
have a similar shape and peak frequency value.
The next complex model was an extension of this idea, with instead of one layer of tilted
fractures, many layers of fracture inclined at the same angle were used to fill the entire
model. This set up resulted in fractures which had a fixed position and orientation. The
7-1. CONCLUSIONS
57
model was then further modified to make the distribution of the fractures random but with
the same angle of 45 degress, and then later to have completely random distributions and
orientations. Going from the one layer case to the multi layered case there is a slight decrease
in amplitudes observed. This is also true when changing from the fixed positions to random
fractures. In general the shape of the spectras remains similar each time and shows a lot of
attenuation.
Following on from this, a model was created where fractures were oriented along two preferential directions at 90 degrees to each other. The one axis was kept at 45 degrees so that it is
comparable to the other models. The presence of 2 perpendicular fracture axis had the effect
of lowering the frequencies signifiicanty compared to the previous models.
The results seen for the parallel fracture set could be explained by the fact that greater clustering of fractures leads to greater energy attenuation [Vlastos et al., 2003]. For the graphs
representing 2 and 5 fractures, there is far less change of shape of the amplitude spectra
because there is less effect on the propagating wavefield. However for a set of 50 fractures
with small normal spacings of less than 10m, there is greater clustering in the model and a
greater attenuation is observed.
In the complex models presented here, the fractures were all longer than the wavelength.
The amplitudes of the reflected wavefield that was recorded at the surface is then a result of
interferance between the many fracture which all act as seperate interfaces.
The results presented from the various models show that even a single fracture has a significant
effect on the propagating wavefield. In agreement with [Boadu and Long, 1997], a single
fracture causes significant reflections and converted waves.
From the reults in both the time and frequency domains, a strong attenuation effect is noticed
for a medium in the presence of fractures. The results in the time domain regarding velocity
changes are poor and the use of S waves did not reveal significant information about veloity
changes due to fractures.
7-1
Conclusions
As stated at the beginning of this report, the aim of this study was to investigate whether
the seismic method is suitable to detect fractures and fracture networks. Another point of
interest was to what extent can the fracture network be identified and characterised.
In the previous discussion the results have been presented and there are some conclusions
which can be made. For a single fracture, many results can be clearly seen such as a clear
orientation effect and length effect.
The analysis between P and S waves led to some very interesting results, with the different
waves proving sensitive to different orientation due to their particle motion.
When the effect of larger numbers of fractures was investigated, some limitations of the
seismic method for fracture characterisation emerged. Although the presence of fractures
seems to significantly lower amplitudes in general, it proved harder to distinguish between
different fracture patterns when large clusters of fractures were considered. Indeed when the
orientations and positions of a group of fractures was changed from completely random to
58
fixed positions very little difference was observed in the resulting data, making it hard to infer
anything about the characteristics.
The results of this work suggest that while fractures can certainly be detected using seismic
methods, they cannot really be identified to any great certainty. As mentioned previously,
fractures most often occur in clusters and groups, but the results here, indicate that these
groups are much harder to characterise than for example a single fracture or fault.
7-2
Future recommendations
Since the results of this study have clearly shown that more realistic and complex fracture
models are hard to identify and characterise, more sophisticated methods seem to be required
for the investigation of realistic fracture scenarios. However, one must realise that the fracture
models used in this study modelled fractures more or less as scatterers. In other words, it
might be required to include more sophisticated fracture theory (different slip conditions)
in the modelling, with appropriate boundary conditions, in order to be able to distinguish
different fractures. In addition, one might consider to investigate the effect of different fracture
fill more thoroughly, for example oil or water fill. In order to do this, different wave types
are needed that are sensitive to different fluids or other fracture parameters. One can think
of combining seismic modelling with electromagnetic methods, or use more complex methods
like poroelastic or seismoelectric wave systems for the numerical modelling.
Bibliography
[Aki and Larner, 1970] Aki, K. and Larner, K. L. (1970). Surface motion of a layered medium
having an irregular interface due to incident plane sh wave. J. Geophys. Res., 75:933.
[Bakulin et al., 2000] Bakulin, A., Grechka, V., and Tsvankin, I. (2000). Estimation of fracture parameters from reflection seismic data - part i: Hti model due to a single fracture
set. Geophysics, 65:1788.
[Boadu and Long, 1997] Boadu, F. K. and Long, L. T. (1997). Effects of fractures on seismicwave velocity and attenuation. Geophysical Journal International, 127:86.
[Hardage, 2011] Hardage, B. (2011). Measuring fracturesquality and quantity. AAPG Explorer, 32:26.
[Hsu and Schoenberg, 1993] Hsu, C. and Schoenberg, M. (1993). Elastic waves through a
simulated fractured medium. Geophysics, 58:964.
[Kearey et al., 2002] Kearey, P., Brooks, M., and Hill, I. (2002). An Introduction to Geophysical Exploration. Blackwell Science.
[King and Xu, 1989] King, M. and Xu, S. (1989). Shear-wave birefringence and directional
permeability in fractured rock. One Petro.
[McKay and Kammer, 1999] McKay, L. D. and Kammer, T. (1999). Incorporating hydrogeology in a mapping-based geology field camp. Journal of Geoscience Education, 47:124.
[Ranalli, 1995] Ranalli, G. (1995). Rheology of the Earth. Chapman and Hall.
[Reynolds, 1997] Reynolds, J. M. (1997). An introduction to applied and environmental geophysics. John Wiley.
[Robertsson et al., 1994] Robertsson, J. O. A., Blanch, J. O., and Symes, W. W. (1994).
Viscoelastic finite-difference modeling. GEOPHYSICS, 59:1444.
[Schoenberg and Douma, 1988] Schoenberg, M. and Douma, J. (1988). Elastic wave propagation in media with parallel fractures and aligned cracks. Geophysical Prospecting, 36:571.
[Sheriff, 1978] Sheriff, R. E. (1978). A first course in geophysical exploration and interpretation. Boston: International Human Resources Development Co.
59
60
Bibliography
[Singhal and Gupta, 1999] Singhal, B. B. S. and Gupta, R. P. (1999). Applied Hydrogeology
of Fractured Rocks. Springer.
[Sudicky, 1989] Sudicky, E. (1989). The laplace transform galerkin technique: A timecontinuous finite element theory and application to mass transport in groundwater. Water
Resources Research, 25:1833.
[Thorbecke, 2013] Thorbecke, J. (2013). Fdelmodc manual. Technical manual.
[Twiss and Moores, 1992] Twiss, R. and Moores, E. (1992). Structural Geology. W. H. Freeman.
[Virieux, 1986] Virieux, J. (1986). P-sv wave propagation in heterogeneous media: Velocitystress finitedifference method. GEOPHYSICS, 51(4):889.
[Vlastos et al., 2003] Vlastos, S., Liu, E., Main, I. G., and Li, X.-Y. (2003). Numerical
simulation of wave propagation in media with discrete distributions of fractures: effects of
fracture sizes and spatial distributions. Geophysical Journal International, 152:649.
[Vlastos et al., 2007] Vlastos, S., Liu, E., Main, I. G., and Narteau, C. (2007). Numerical
simulation of wave propagation in 2-d fractured media: scattering attenuation at different
stages of the growth of a fracture population. Geophysical Journal International, 171:865.
[Wapenaar and Berkhout, 1989] Wapenaar, C. and Berkhout, A. (1989). Elastic Wavefield
Extrapolation: Redatuming of Single- and Multi-component seismic data. Elsevier.