Você está na página 1de 236

11 Reservoir Geophysics

Introduction Elastic Waves and Rock Properties Seismic Resolution Vertical Resolution Lateral
Resolution Analysis of Amplitude Variation with Oset Reection and Refraction Reector Curvature
AVO Equations Processing Sequence for AVO Analysis Derivation of AVO Attributes by Prestack Amplitude
Inversion Interpretation of AVO Attributes 3-D AVO Analysis Acoustic Impedance Estimation Synthetic
Sonic Logs Processing Sequence for Acoustic Impedance Estimation Derivation of Acoustic Impedance Attribute
3-D Acoustic Impedance Estimation Instantaneous Attributes Vertical Seismic Proling VSP Acquisition
Geometry Processing of VSP Data VSP-CDP Transform 4-D Seismic Method Processing of 4-D Seismic
Data Seismic Reservoir Monitoring 4-C Seismic Method Recording of 4-C Seismic Data Gaisers Coupling
Analysis of Geophone Data Processing of P P Data Rotation of Horizontal Geophone Components CommonConversion-Point Binning Velocity Analysis of P S Data Dip-Moveout Correction of P S Data Migration of P S
Data Seismic Anisotropy Anisotropic Velocity Analysis Anisotropic Dip-Moveout Correction Anisotropic
Migration Eect of Anisotropy on AVO Shear-Wave Splitting in Anisotropic Media Exercises Appendix L:
Mathematical Foundation of Elastic Wave Propagation Stress-Strain Relation Elastic Wave Equation
Seismic Wave Types Body Waves and Surface Waves Wave Propagation Phenomena Diraction, Reection,
and Refraction The Zoeppritz Equations Prestack Amplitude Inversion References

11.0 INTRODUCTION
In Chapter 8, we reviewed the two-dimensional (2-D)
and three-dimensional (3-D), post- and prestack migration strategies for imaging the earths interior in depth.
In Chapter 9, we learned traveltime inversion techniques
for estimating a structural model of the earth that is
needed to obtain an accurate image in depth. In Chapter 10, structural inversion case studies for earth modeling and imaging in depth were presented. By structural
inversion, we dene the geometry of the reservoir unit,
and the overlying and underlying depositional units.
Traveltimes, however, are only one of the two compo-

nents of recorded seismic waveelds; amplitudes are the


other component.
In this chapter, we shall turn our attention to inversion of reection amplitudes to infer petrophysical
properties within the depositional unit associated with
the reservoir rocks. The petrophysical properties include porosity, permeability, pore pressure, and uid
saturation. Specically, we shall discuss prestack amplitude inversion to derive the amplitude variation with
oset (AVO) attributes (Section 11.2) and poststack
amplitude inversion to estimate an acoustic impedance

1794

Seismic Data Analysis

model of the earth (Section 11.3). The processes of estimating the acoustic impedance and AVO attributes
by way of inversion of amplitudes may be appropriately referred to as stratigraphic inversion. Our goal
ultimately is reservoir characterization based on structural and stratigraphic inversion of seismic data with
calibration to well data.
We appropriately begin this chapter by investigating the resolution we can achieve from seismic data in
dening vertical and lateral variations in the geometry
of the reservoir unit. Resolution is the ability to separate two events that are very close together (Section
11.1). There are two aspects of seismic resolution: vertical (or temporal) and lateral (or spatial). Seismic resolution becomes especially important in mapping small
structural features, such as subtle sealing faults, and in
delineating thin stratigraphic features that may have
limited areal extent.
Reservoir characterization involves calibration of
the results of analysis of surface seismic data both
from structural and stratigraphic inversion, to well
data. One category of well data includes various types
of logs recorded in the borehole. Logs that are most
relevant to seismic data are sonic, shear, and density.
Another category of well data is a vertical seismic prole (VSP) (Section 11.4).
Just as we can seismically characterize a reservoir,
we also can seismically monitor its depletion. This is
achieved by recording 3-D seismic data over the eld
that is being developed and produced at appropriate
time intervals and detecting changes in the reservoir
conditions; specically, changes in petrophysical properties of the reservoir rocks, such as uid saturation and
pore pressure. Specically, such changes may be related
to changes in the seismic amplitudes from one 3-D survey to the next. Time-lapse 3-D seismic monitoring of
reservoirs is referred to as the 4-D seismic method (Section 11.5). The fourth dimension represents the calendar
time over which the reservoir is being monitored.
Some reservoirs can be better identied and monitored by using shear-wave data. For instance, acoustic
impedance contrast at the top-reservoir boundary may
be too small to detect, whereas shear-wave impedance
contrast may be suciently large to detect. By recording multicomponent data at the ocean bottom, P -wave
and S-wave images can be derived. Commonly, four
data components are recorded the pressure waveeld,
and inline, crossline, and vertical components of particle
velocity. Thus, the multicomponent seismic data recording and analysis is often referred to as the 4-C seismic
method (Section 11.6).
This chapter ends with a brief discussion on
anisotropy. While exploration seismology at large is
based on the assumption of an isotropic medium, the

earth in reality is anisotropic. This means that elastic


properties of the earth vary from one recording direction to another. Seismic anisotropy often is associated
with directional variations in velocities. For instance,
in a vertically fractured limestone reservoir, velocity in
the fracture direction is lower than velocity in the direction perpendicular to the plane of fracturing (azimuthal
anisotropy). Another directional variation of velocities
involves horizontal layering and fracturing of rocks parallel to the layering. In this case, velocity in the horizontal direction is higher than the vertical direction
(transverse isotropy). In Section 11.7, we shall review
seismic anisotropy in relation to velocity analysis, migration, DMO correction, and AVO analysis.

Elastic Waves and Rock Properties


Seismic waves induce elastic deformation along the
propagation path in the subsurface. The term elastic
refers to the type of deformation that vanishes upon removal of the stress which has caused it. To study seismic
amplitudes and thus investigate their use in exploration
seismology, it is imperative that we review wave propagation in elastic solids. This gives us the opportunity
to appreciate the underlying assumptions in estimating
acoustic impedance and AVO attributes.
A summary of the elastic wave propagation theory
is provided in Appendix L. To facilitate the forthcoming
discussion on the link between elastic waves and rock
properties, rst, we summarize the denitions of elastic
wave theory that should always be remembered.
(a) Stress is force per unit area. Imagine a particle represented by an innitesimally small volume
around a point within a solid body with dimensions
(dx, dy, dz) as depicted in Figure L-1. The stress
acting upon one of the surfaces, say dy dz, can
in general be at some arbitrary direction. It can,
however, be decomposed into three components
one which is normal to the surface, and two which
are tangential to the surface. The normal component of the stress is called the normal stress and the
tangential components are called the shear stress.
A normal stress component is tensional if it is positive and compressional if it is negative. Fluids cannot support shear stress. In a uid medium, only
one independent stress component exists the hydrostatic pressure.
(b) Strain is deformation measured as the fractional
change in dimension or volume induced by stress.
Strain is a dimensionless quantity. The stress eld
away from the typical seismic source is so small

Reservoir Geophysics

(c)
(d)
(e)

(f)
(g)

(h)

that it does not cause any permanent deformation on rock particles along the propagation path.
Hence, the strain induced by seismic waves is very
small, usually around 106 . Consider two points, P
and Q, within a solid body as indicated in Figure
L-2. Subject to a stress eld, the solid is deformed
in some manner and the particles at points P and
Q are displaced to new locations P and Q . Consider deformations of specic types illustrated in
Figure L-3. The simplest deformation is the extension in one direction as a result of a tensional stress
(Figure L-3a). The fractional change in length in
a given direction is dened as the principal strain
component. A positive strain refers to an extension and a negative strain refers to a contraction.
Other types of deformation are caused by shearing
(Figure L-3b), rotation (Figure L-3c), and a combination of the two (Figure L-3d). These angular
deformations are called shear strains since they result in a shearing of the volume around a point
within a solid body (Figure L-3b).
Elastic deformation is a deformation in solid bodies
that vanishes once the stress is released.
Hookes law for elastic deformations states that the
strain at any point is directly proportional to the
stresses applied at that point.
Elastic moduli are material constants that describe
stress-strain relations:
(1) Bulk modulus is the ratio of hydrostatic stress
to volumetric strain; hence, it is a measure of incompressibility.
(2) Modulus of rigidity is the ratio of shear stress
to shear strain; hence, it is a measure of resistance
to shear stress.
(3) Youngs modulus is the ratio of the longitudinal
stress to the longitudinal strain associated with a
cylindrical rod that is subjected to a longitudinal
extension in the axial direction. Since strain is a
dimensionless quantity, Youngs modulus has the
dimensions of stress.
(4) Poissons ratio is the ratio of the lateral contraction to longitudinal extension associated with
a cylindrical rod that is subjected to a longitudinal extension in the axial direction. Since strain is
a dimensionless quantity, Poissons ratio is a pure
number.
Seismic waves are elastic waves that propagate in
the earth.
P -waves (or equvalently, compressional waves, longitudinal waves, or diltatational waves) are waves
with particle motion in the direction of wave propagation.
S-waves (or equivalently, shear waves, transverse
waves, or rotational waves) are waves with particle

1795

motion in the direction perpendicular to the direction of wave propagation.


(i) Reection is the waveeld phenomenon associated
with the fraction of incident wave energy that is
returned from an interface that separates two layers
with dierent elastic moduli.
(j) Refraction is the the waveeld phenomenon associated with the fraction of incident wave energy that
is transmitted into the next layer.
(k) Diraction is the waveeld phenomenon associated
with energy that propagates outward from a sharp
discontinuity in the subsurface.
In exploration seismology, we are primarily interested in compressional and shear waves that travel
through the interior of solid layers, and thus are characterized as body waves. Whereas in earthquake seismology, we also make use of Love and Rayleigh waves which
travel along layer boundaries, and thus are characterized as surface waves.
Both body waves and surface waves are dierent
forms of elastic waves, each associated with a specic
type of particle motion. In the case of compressional
waves, the particle motion induced by a compressional
stress is in the direction of wave propagation. The compressional stress causes a change in the particle dimension or volume. The more the rock resists to the compressional stress, the higher the compressional wave velocity. In the case of shear waves, the particle motion
induced by a shear stress is in the direction perpendicular to the direction of wave propagation. The shear
stress does not cause a change in the particle dimension
or volume; instead, it changes particle shape. The more
the rock resists shear stress, the higher the shear wave
velocity. Under the assumption that both wave types
are elastic, whatever the change induced by the wave
motion the elastic deformation in particle shape, dimension or volume, vanishes once the wave motion on
the particle vanishes and is propagated onto the neighboring particle.
Figure 11.0-1 outlines the interrelationships between the various elastic parameters. Starting with
Youngs modulus the ratio of principal stress to principal strain, and Poissons ratio the ratio of shear
strain to principal strain, Lames constants and
are dened. The Lame constant is indeed the modulus
of rigidity and the Lame constant = (2/3), where
is the bulk modulus (Section L.3). Then, the two
wave velocities compressional (P -waves) and shear
(S-waves), are derived in terms of the Lame constants,
or the bulk modulus and modulus of rigidity, and density.
From the denitions of the P - and S-wave velocities
in Figure 11.0-1, note that both are inversely proportional to density . At rst thought, this means that

1796

Seismic Data Analysis

FIG. 11.0-1. The relationships between the various elastic parameters for isotropic solids.

Reservoir Geophysics

FIG. 11.0-2. Variations in P -wave velocity with various rock types with dierent densities (Gardner et al., 1974).

1797

1798

Seismic Data Analysis

FIG. 11.0-3. Crossplot of P -wave slowness versus S-wave slowness based on laboratory measurements wtih various rock
types (Pickett, 1963).

the lower the rock density the higher the wave velocity. A good example is halite which has low density (1.8
gr/cm3 ) and high P -wave velocity (4500 m/s). In most
cases, however, the higher the density the higher the
velocity (Figure 11.0-2). This is because an increase in
density usually is accompanied by an increase in the
ability of the rock to resist compressional and shear
stresses. So an increase in density usually implies an
increase in bulk modulus and modulus of rigidity. Returning to the expressions for the P - and S-wave velocities in Figure 11.0-1, note that the greater the bulk
modulus or the modulus of rigidity, the higher the velocity. Based on eld and laboratory measurements, Gardner et al. (1974) established an empirical relationship
between density and P -wave velocity . Known as
Gardners formula for density, this relationship given
by = c0.25 , where c is a constant that depends on
the rock type, is useful to estimate density from velocity when the former is unknown. With the exception of
anhydrites, most rock types sandstones, shales, and
carbonates, tend to obey Gardners equation for density.

In Section 3.0, a brief review of the results of some


of the key laboratory experiments on seismic velocities was made. For a given lithologic composition, seismic velocities in rocks are inuenced by porosity, pore
shape, pore pressure, pore uid saturation, conning
pressure, and temperature. It is generally accepted that
conning pressure, and thus the depth of burial, has the
most profound eect on seismic velocities (Figure 3.03). For instance, the P -wave velocities for clastics can
vary from 2 km/s at the surface up to 5 km/s and for
carbonates from 3 km/s at the surface up to 6 km/s at
depths greater than 5 km.
Because of the large variations in P -wave velocities caused by all these factors, P -wave velocity alone is
not adequate to infer the lithology, unambiguously. The
ambiguity in lithologic identication can be resolved to
some extent if we have the additional knowledge of Swave velocities. Here, we examine the ratio of the Swave velocity to the P -wave velocity, /, which only
depends on Poissons ratio (Figure 11.0-1). In some
instances, we refer to the inverse ratio /. The higher
the Poissons ratio, the higher the velocity ratio /.

Reservoir Geophysics

1799

FIG. 11.0-4. (a) Crossplot of P -wave velocity versus S-wave velocity derived from full-waveform sonic logs using rock samples
with dierent lithologies SS: sandstone, SH: shale and LS: limestone; (b) crossplot of the velocity ratio versus the P -wave
velocity using the same sample points as in (a) (Miller and Stewart, 1999).

1800

Seismic Data Analysis

FIG. 11.0-5. Variation of the velocity ratio with respect to (a) shale content and (b) clay content (Miller and Stewart, 1999).

Reservoir Geophysics
This relationship is supported by the physical meaning
of Poissons ratio the ratio of shear strain to principle
strain. A way to describe the physical meaning of Poissons ratio is to consider a metal rod that is subject to
an extensional strain. As the rod is stretched, its length
increases while its thickness decreases. Hence, the less
rigid the rock, the higher the Poissons ratio. This is exactly what is implied by the expression in Figure 11.0-1
that relates the modulus of rigidity to Poissons ratio
. Unconsolidated sediments or uid-saturated reservoir
rocks have low rigidity, hence high Poissons ratio and
high velocity ratio /. Here is the rst encounter with
a direct hydrocarbon indicator the P - to S-wave velocity ratio. Ostrander (1984) was the rst to publish
the link between a change in Poissons ratio and change
in reection amplitude as a function of oset.
Aside from the direct measurement of S-wave velocities down the borehole, there are three indirect ways
to estimate the S-wave velocities. The rst approach
is to perform prestack amplitude inversion to estimate
the P - and S-wave reectivities and thus compute the
corresponding acoustic impedances (Section 11.2). The
second approach is to record multicomponent seismic
data and estimate the S-wave velocities from the P -toS converted-wave component (Section 11.6). The third
approach is to generate and record S-waves themselves.
Figure 11.0-3 shows a plot of the S-wave slowness (inverse of the S-wave velocity) versus the P -wave
slowness (inverse of the P -wave velocity) based on loboratory measurements (Pickett, 1963). Figure 11.0-4a
shows a plot of the P -wave velocity to S-wave velocity
based on full-waveform sonic logs from a producing oil
eld (Miller and Stewart, 1999). The key observation
made from these results is that a lithologic composition
may be associated with a reasonably distinctive velocity ratio /. The shale and limestone samples fall on
a linear trend that corresponds to a velocity ratio of
1.9, whereas the dolomite samples have a velocity ratio
of 1.8. The sandstone samples have a range of velocity
ratio of 1.6 to 1.7. Lithologic distinction sometimes is
more successful with a crossplot of P -wave to S-wave
velocity ratio versus the P -wave velocity itself (Miller
and Stewart, 1999). This is illustrated in Figure 11.04b which shows the same sample points as in Figure
11.0-4a.
Eect of shale and clay content on the velocity ratio
/ is an important factor in lithologic identication.
Field and laboratory data from sandstone cores indicate
that the velocity ratio / increases with increasing
shale and clay content as a result of a decrease in Swave velocity (Figure 11.0-5).
Finally, eect of porosity on the velocity ratio /
is generally dictated by the pore shape. For limestones
with their pores in the form of microcracks, the velocity ratio increases as the percent porosity increases

1801

(Eastwood and Castagna, 1983). For sandstones with


their rounded pores, the velocity ratio does not increase
as much with increasing porosity (Miller and Stewart,
1999).The dierence between the rounded pores and microcracks lies in the fact that it is easier to collapse a
rock with microcracks, hence lower modulus of rigidity.

11.1 SEISMIC RESOLUTION


Resolution relates to how close two points can be, yet
still be distinguished. Two types of resolution are considered vertical and lateral, both of which are controlled by signal bandwidth. The yardstick for vertical
resolution is the dominant wavelength, which is wave
velocity divided by dominant frequency. Deconvolution
tries to increase the vertical resolution by broadening
the spectrum, thereby compressing the seismic wavelet.
The yardstick for lateral resolution is the Fresnel zone,
a circular area on a reector whose size depends on
the depth to the reector, the velocity above the reector and, again, the dominant frequency. Migration
improves the lateral resolution by decreasing the width
of the Fresnel zone, thus separating features that are
blurred in the lateral direction.

Vertical Resolution
For two reections, one from the top and one from the
bottom of a thin layer, there is a limit on how close they
can be, yet still be separable. This limit depends on the
thickness of the layer and is the essence of the problem
of vertical resolution.
The dominant wavelength of seismic waves is given
by
v
(11 1)
= ,
f
where v is velocity and f is the dominant frequency.
Seismic wave velocities in the subsurface range between
2000 and 5000 m/s and generally increase in depth. On
the other hand, the dominant frequency of the seismic
signal typically varies between 50 and 20 Hz and decreases in depth. Therefore, typical seismic wavelengths
range from 40 to 250 m and generally increase with
depth. Since wavelength determines resolution, deep
features must be thicker than the shallow features to be
resolvable. A graph of wavelength as a function of velocity for various values of frequency is plotted in Figure
11.1-1. The wavelength is easily determined from this
graph, given the velocity and dominant frequency.

1802

Seismic Data Analysis

Table 11-1. Threshold for vertical resolution.


/4 = v/4f
v (m/s)

f (Hz)

/4 (m)

2000
3000
4000
5000

50
40
30
20

10
18
33
62

The acceptable threshold for vertical resolution


generally is a quarter of the dominant wavelength. This
is subjective and depends on the noise level in the
data. Sometimes the quarter-wavelength criterion is too
generous, particularly when the reection coecient is
small and no reection event is discernable. Sometimes
the criterion may be too stringent, particularly when
events do exist and their amplitudes can be picked with
ease.
Table 11-1 contains the wavelength threshold values for vertical resolution, considering the realistic velocity and frequency ranges. For example, a shallow feature with a 2000-m/s velocity and 50-Hz dominant frequency potentially can be resolved if it is as thin as 10
m. A thinner feature cannot be resolved. Similarly, for
a deep feature with a velocity as high as 5000 m/s and
dominant frequency as low as 20 Hz, the thickness must
be at least 62 m for it to be resolvable.
It is now appropriate to ask whether a thin stratigraphic unit must be resolved to be mapped. The answer is no. Resolution as dened here and in the geophysical literature implies that reections from the top
and bottom of a thin bed are seen as separate events
or wavelet lobes. Using this denition, resolution does
not consider amplitude eects. The thickness and areal
extent of beds below the resolution limit often can
be mapped on the basis of amplitude changes. This
amplitude-based analysis can be especially precise when
used for mapping gas-generated bright spots in tertiary
rocks. Thus, in many stratigraphic plays, resolution in
the strict sense is not an issue. For these plays, detection, not resolution, is the problem.
Vertical resolution is a concern when discontinuities are inferred along a reection horizon because of
faults. Figure 11.1-2 shows a series of faults with vertical
throws that are equal to 1, 1/2, 1/4, 1/8, and 1/16th
of the dominant wavelength. It is when the throw is
equal or greater then one-fourth of the dominant wavelength that the presence of the fault can be inferred
easily. Perhaps a smaller throw can be inferred by using diractions from faults along the reection horizon,
provided the noise level is low in the data.

FIG. 11.1-1. The relationship between velocity, dominant frequency, and wavelength. Here, wavelength = velocity/frequency. (Adapted from Sheri, 1976; courtesy American Association of Petroleum Geologists.)

Clearly the ability to resolve or detect small targets can be increased by increasing the dominant frequency of the stacked data. The dominant frequency of
a stacked section from a given area is governed by the
physical properties of the subsurface, processing quality,
and recording parameters. Since we cannot control the
subsurface properties, the high-frequency signal level
can only be inuenced by the eort put into recording
and processing.
The emphasis in recording should be to preserve
high frequencies and suppress noise. The sampling rate
and antialiasing lters should be adequate to record
the desired frequencies. Receiver arrays should be small
enough to prevent the signicant loss of high-frequency
signal because of intragroup moveout and statics. However, the arrays should not be too small, since small
arrays are not as eective at suppressing random, highfrequency noise (wind noise) as large arrays. Finally, the
source eort should be high enough to provide adequate
signal level relative to noise level within the desired frequency band. Unless the signal-to-noise ratio of the eld
data is above some minimal level, say 0.25, processing
algorithms have diculty in recovering the signal. The
signal has to be detectable before it can be enhanced.
The emphasis in processing should be to preserve
and display the high-frequency signal present in the
input data. Filters with good high-frequency response
should be used for interpolation processes such as NMO

Reservoir Geophysics

1803

FIG. 11.1-2. Faults with dierent amounts of vertical throws expressed in fractions of the dominant wavelength.

removal, datum and statics corrections, and multiplex


skew corrections. Extra care should be taken to ensure that small-scale residual moveout or statics, which
might cause loss of high-frequency signal during stacking, are removed before stack. Nonsurface-consistent
alignment programs (often called trim statics programs)
sometimes are used for this purpose. Finally, care must
be taken to ensure that all the high-frequency signal is
displayed on the nal stack. Poststack deconvolution is
a useful tool for this purpose.

Lateral Resolution
Lateral resolution refers to how close two reecting
points can be situated horizontally, yet be recognized
as two separate points rather than one. Consider the
spherical wavefront that impinges on the horizontal planar reector AA in Figure 11.1-3. This reector can
be visualized as a continuum of point diractors. For
a coincident source and receiver at the earths surface
(location S), the energy from the subsurface point (0)
arrives at t0 = 2z0 /v. Now let the incident wavefront
advance in depth by the amount /4. Energy from subsurface location A, or A, will reach the receiver at time
t1 = 2(z0 + /4)/v. The energy from all the points
within the reecting disk with radius OA will arrive
sometime between t0 and t1 . The total energy arriving
within the time interval (t1 t0 ), which equals half the
dominant period (T /2), interferes constructively. The
reecting disk AA is called a half-wavelength Fresnel
zone (Hilterman, 1982) or the rst Fresnel zone (Sheri,
1991). Two reecting points that fall within this zone
generally are considered indistinguishable as observed
from the earths surface.

Since the Fresnel zone depends on wavelength, it


also depends on frequency. For example, if the seismic
signal riding along the wavefront is at a relatively high
frequency, then the Fresnel zone is relatively narrow.
The smaller the Fresnel zones, the easier it is to dierentiate between two reecting points. Hence, the Fresnelzone width is a measure of lateral resolution. Besides
frequency, lateral resolution also depends on velocity
and the depth of the reecting interface the radius
of the wavefront is approximated by (Exercise 11-1)
r=

z0
.
2

(11 2a)

In terms of dominant frequency f (equation 11-1), the


Fresnel-zone width is

FIG. 11.1-3. Denition of the Fresnel zone AA .

1804

Seismic Data Analysis

FIG. 11.1-4. A constant-velocity zero-oset section from an earth model that consists of four reectors, each with four
nonreecting segments A, B, C, and D. Lateral resolution is governed by the size of the Fresnel zone. The lateral extent of
each gap is indicated by the solid bars on top. Note that A is hardly recognizable on any of the four horizons; B can be inferred
on the shallow horizon at 0.5 s; C is dicult to infer after 2 s, while D is recognizable at all depths. All of these observations
depend on noise level and how easily the diractions can be recognized.

Reservoir Geophysics
Table 11-2. Threshold for lateral resolution (rst Fresnel zone).
r = (v/2) t0 /f
t0 (s)

v (m/s)

f (Hz)

r (m)

1
2
3
4

2000
3000
4000
5000

50
40
30
20

141
335
632
1118

r=

v
2

t0
.
f

(11 2b)

Table 11-2 shows the Fresnel zone radius, where


r = OA in Figure 11.1-3 for a range of frequency
and velocity combinations at various depths t0 = 2z/v.
From Table 11-2, note that the shallower the event
(and the higher the dominant frequency), the smaller
the Fresnel zone. Since the Fresnel zone generally increases with depth, spatial resolution also deteriorates
with depth.
Figure 11.1-4 shows reections from four interfaces,
each with four nonreecting segments. The actual sizes
of these segments are indicated by the solid bars on
top. On the seismic section, the reections appear to
be continuous across some of these segments. This is
because the size of some of the nonreecting segments
is much less than the width of the Fresnel zone; they
are beyond the lateral resolution limit.
Spatial resolution is better understood in terms
of diractions. Note that in Figure 11.1-4, the diraction energy is smeared across the nonreecting segments
on the deeper reectors. Since migration is the process
that collapses diractions, it is reasonable to think that
migration increases spatial resolution. Remember that
migration can be achieved by downward continuation
of receivers from the surface to the reecting horizons.
As a result of downward continuation, the observation
points get closer and closer to the reection points and,
therefore, the Fresnel zone gets smaller and smaller. A
smaller Fresnel zone means a higher spatial resolution
(equation 11-2).
Migration tends to collapse the Fresnel zone to approximately the dominant wavelength (equation 11-1)
(Stolt and Benson, 1986). Therefore, we anticipate that
migration will not resolve the horizontal limits of some
of the nonreecting segments along the deeper reectors
in Figure 11.1-4. Tables 11-1 and 11-2 can be used to estimate the potential resolution improvement that may
result from migration. Unless three-dimensional (3-D)
migration (Section 7.3) is performed, the actual resolution will be less than that indicated. Two-dimensional

1805

(2-D) migration only shortens the Fresnel zone in the


direction parallel to the line orientation. Resolution in
the perpendicular direction is not aected by 2-D migration.
Figure 11.1-5 indicates how vertical and lateral resolution problems are inter-related. We want to determine the edge of the pinchout. The basis of the pinchout
model is a wedge of material represented at a given midpoint location by a two-term reectivity sequence, one
term associated with the top and one with the bottom
of the wedge. The true thickness of the wedge at various locations is indicated on top of Figure 11.1-5a. The
velocity within the wedge is 2500 m/s.
We rst consider the reectivity sequence with
two spikes of equal amplitude and identical polarity.
The vertical-incidence seismic response (Figure 11.15a) is obtained by convolving the sequences with a
zero-phase wavelet with a 20-Hz dominant frequency.
(The zero-phase response simplies event tracking from
the top and bottom of the wedge.) Based on this response, the edge of the wedge can be inferred as left
of location B, where the waveform reduces to a single
wavelet (Figure 11.1-5a). From the resolution threshold criterion, the smallest thickness that can be resolved is (2500 m/s)/(4Hz) = 31.25 m. Figures 11.15a, b, and c show the same pinchout modeled using three dierent zero-phase wavelets with increasing dominant frequency (20, 30, and 40 Hz). Separation between the true location of pinchout A and
the position of the minimum resolvable wedge thickness B decreases with increasing wavelet bandwidth.
While the resolution threshold criterion allows us
to say only that the thickness of the wedge is less than
31.25 m left of location B, an amplitude-based criterion can provide a substantially more accurate location for the edge of the wedge. Again, refer to Figure
11.1-5a and observe the sudden change in amplitude
at location A where the true edge of the pinchout is
located. Hence, the edge still can be reliably detected,
even though it may not be resolved, provided the signalto-noise (signal-to-noise) ratio is favorable. Assuming
that the relative size of the top and bottom reection
coecients is known, amplitudes also can be used to estimate the wedge thickness between locations B and A.
Figures 11.1-5a, b, and c show an apparent lateral variation in layer thickness. To see the dierence
between the true thickness and the apparent thickness
(peak-to-peak time), refer to Figure 11.1-5d. This gure
shows the data of Figure 11.1-5b with the actual geometry of the wedge superimposed on the seismic response.
Since the composite wavelet has only one positive peak,
the apparent thickness between locations A and B is
nearly zero. At location B, the composite wavelet has a

1806

Seismic Data Analysis

FIG. 11.1-5. (a) The result of convolving a zero-phase wavelet of 20-Hz dominant frequency with a wedge reectivity model.
The reection coecients associated with the top and bottom of the wedge are of equal amplitude and identical polarity. The
true edge of the wedge is beneath location A and the true thickness of the wedge is indicated by the numbers on top; (b)
same as (a) except the dominant frequency of the wavelet is 30 Hz; (c) same as (a) except the dominant frequency of the
wavelet is 40 Hz; (d) same as (b) with the actual geometry of the wedge superimposed on the seismic response; (e) same as
(b) except the reection coecients from the top and bottom of the wedge have opposite polarity; (f) same as (e) with the
actual geometry of the wedge superimposed on the seismic response.

at top. Immediately to the right of location B, the at


top disappears and the composite wavelet splits. The
at-top character can be identied as the limit of vertical resolution (Ricker, 1953). A short distance to the
right of the point at which the composite wavelet rst
splits into two peaks, the apparent thickness becomes
equal to the true thickness. This thickness is called the
tuning thickness and is equal to peak-to-trough separation (one half the dominant period) of the convolving
wavelet (Kallweit and Wood, 1982). Beyond the point
of tuning thickness, note the apparent thickening of the
layer between locations B and C. To the right of location C, the apparent and true thicknesses become equal.

Besides apparent thickness, the maximum absolute


amplitude of the composite wavelet along the pinchout
also changes (Kallweit and Wood, 1982). To the left of
location A in Figure 11.1-5b, note the single isolated
zero-phase wavelet. Immediately to the right of location A, the response of the two closely spaced spikes
with identical polarity results in the maximum absolute amplitude. This amplitude gradually decreases to
a minimum exactly at the tuning thickness. It then increases and reaches the amplitude value of the original
single wavelet to the right of location C.
Maximum amplitude and apparent thickness
change in reverse when the reectivity model consists

Reservoir Geophysics

1807

thicknesses become equal. Immediately to the right of


location A in Figure 11.1-5e, the response of the two
closely spaced spikes with opposite polarity results in
the cancellation of the amplitudes. The largest absolute
amplitude of the composite wavelet gradually increases
to a maximum immediately to the right of location B.
It gradually decreases and reaches the amplitude value
of the original single wavelet to the right of location C.
From the above discussion, we see that peak-topeak time measurements and amplitude information
can aid in detecting pinchouts that may otherwise be
unresolvable. If the size of the reection coecients were
known, then the amplitudes could be used to map the
thickness below the resolution limit.
Nevertheless, the reliability of the analysis depends
to some extent on the signal-to-noise ratio. Finally, the
deceptive changes in amplitude and apparent thickness
must be noted during the mapping of the top and bottom of the pinchout.
11.2 ANALYSIS OF AMPLITUDE
VARIATION WITH OFFSET

FIG. 11.2-1. A moveout-corrected CMP gather with a reection event at 1.25 s that exhibits amplitude variations
with oset. (Courtesy Western Geophysical.)

of reection coecients with equal amplitude and opposite polarity (Figure 11.1-5e). The composite waveform resulting from this reectivity model is discussed
by Widess (1973). Two spikes of opposite polarity with
a small separation between them act as a derivative operator. When applied to a zero-phase wavelet, this operator causes a 90-degree phase shift. This phase shift
can be seen in Figure 11.1-5e on the wavelets between
locations A and B. Widess (1973) observed that the
composite wavelet within this zone basically retains its
shape while its amplitude changes.
Figure 11.1-5f shows data of Figure 11.1-5e with
the actual geometry of the wedge superimposed on the
seismic response. Note that the wedge appears thicker
than it actually is between locations A and B. Also note
the apparent thinning of the layer between locations
B and C. Beyond location C, the apparent and true

In Appendix L, we review the theory of seismic wave


propagation in an elastic continuum. The earths upper
crust that is of interest in seismic prospecting, however, is made up of rock layers with dierent elastic
moduli. When seismic waves travel down in the earth
and encounter layer boundaries with velocity and density contrast, the energy of the incident wave is partitioned at each boundary. Specically, part of the incident energy associated with a compressional source
is mode-converted to a shear wave; then, both the
compressional- and shear-wave energy are partly reected from and partly transmitted through each of
these layer boundaries.
The fraction of the incident energy that is reected
depends upon the angle of incidence. Analysis of reection amplitudes as a function of incidence angle can
sometimes be used to detect lateral changes in elastic
properties of reservoir rocks, such as change in Poissons
ratio. This may then suggest a change in the ratio of P wave velocity to S-wave velocity, which in turn may
imply a change in uid saturation within the reservoir
rocks.
By way of CMP recording geometry, reection
amplitudes are not measured as a function of angle;
instead, they are measured as a function of sourcereceiver oset. Nevertheless, a range of incidence angles is spanned by a range of osets. Amplitude-versusoset analysis, therefore, provides the information on
amplitude-versus-angle.
Figure 11.2-1 shows a moveout-corrected CMP
gather with a strong reection at 1.25 s. Note the am-

1808

Seismic Data Analysis

plitude variations with oset in this case, amplitudes


increasing with oset. By picking the peak amplitudes
and plotting them against oset, an amplitude variation
with oset (AVO) curve is derived for a target horizon at
each CMP location. Then, an AVO pattern may emerge,
which can then be used to infer reservoir parameters.
The pattern with which amplitudes vary with oset
depends on the combination of reservoir rock and uid
properties. Detection of a pattern is dictated primarily
by the signal-to-noise ratio and the range of incidence
angle that is spanned by the oset range of the CMP
gather for a target horizon. The shallower the reector, the wider the range of incidence angle; hence, AVO
indicators are best determined for shallow targets.
The following discussion on reection and refraction of seismic waves is based on at layer boundaries.
Reection amplitudes also depend upon the dip of the
reecting boundary and its curvature. We can remove
the dip and curvature eects by performing prestack
time migration. The resulting CMP gathers are associated with reectors in their migrated positions and
reection amplitudes can then be associated with a locally at earth model.

For simplicity, consider a monochromatic compressional


plane wave that impinges at normal incidence upon a
at layer boundary at z = 0. The incident energy is partitioned between a reected and transmitted compressional plane wave. For this special case, there is only
one stress component, Pzz , and one displacement component, w, which is only a function of z. The equation
of wave motion (L-29c) for this special case takes the
form
2

w
w
= ( + 2) 2 ,
(11 3a)
2
t
z
where is density of the medium, and and are
Lames constants (equations L-19a,b) associated with
an isotropic solid. They are directly related to the
compressional-wave velocity by equation (L-35) which
is rewritten below as

and for the transmitted plane wave:

w2 (z, t) = A2 exp i z it ,
2

(11 4c)

where 2 is the compressional-wave velocity in the lower


layer.
Given the incident wave amplitude A0 , we want to
compute the reected and refracted wave amplitudes,
A1 and A2 , respectively. Equations (11-4a,b,c) must be
accompanied with boundary conditions at z = 0 that
satisfy the continuity of displacement and stress. The
continuity of displacement condition, w0 + w1 = w2 at
the interface z = 0 gives the relation
A0 + A1 = A2 .

(11 5)

The stress component Pzz can be specialized from


Hookes law (equation L-18c) for the present case of
a normal-incident compressional plane wave
w
.
(11 6)
z
The continuity of stress condition at the interface z = 0
gives the relation
Pzz = ( + 2)

Reection and Refraction

frequency, 1 is the compressional-wave velocity in the


upper layer (equation 11-3b), and z-axis is downward
positive. Similar wave functions can be written for the
reected plane wave:

w1 (z, t) = A1 exp i z it
(11 4b)
1

+ 2
.

(11 3b)

A solution to equation (11-3a) can be written as

w0 (z, t) = A0 exp i z it ,
(11 4a)
1
where w0 is the wave function for the incident compressional wave, A0 is its amplitude, is the angular

Pzz

+ Pzz

= Pzz 2 .

(11 7)

Now, dierentiate the wave functions of equations (114a,b,c) with respect to z, substitute into equation (11-7)
and set z = 0 to obtain the following expression:
1 1 A0 1 1 A1 = 2 2 A2 ,

(11 8)

where 1 and 2 are the densities of the upper and lower


layers, respectively.
We now have two equations, (11-5) and (11-8), and
two unknowns, A1 and A2 . By combining equations (115) and (11-8), we can derive the ratio of the reected
wave amplitude to the incident wave amplitude, which
is called the reection coecient c, associated with the
layer boundary as
c=

A1
2 2 1 1
=
.
A0
2 2 + 1 1

(11 9)

Dene the product of density and velocity as the


seismic impedance, I = . If there is a dierence between the seismic impedances of the two layers, then a
reection occurs at the interface. If the upper layer has
a higher impedance than the lower layer, the reection
coecient becomes negative causing a phase reversal on
the reected waveform.

Reservoir Geophysics

1809

FIG. 11.2-2. A long, deep-water seismic section that shows internal reections within the water layer caused by density
contrast associated with temperature and salinity changes in the water layer. (Data courtesy IFP.)

An impedance contrast at a layer boundary often is


largely caused by velocity contrast. Nevertheless, conditions exist for which density contrast can be signicant in giving rise to reections. Figure 11.2-2 illustrates
one such case. The internal reections within the water
layer occur because of changes in water temperature
and salinity that causes variations in water density.
A typical reection coecient for a strong reector
is about 0.2. The reection coecient associated with
a hard water bottom is about 0.3. Note that it is the
impedance contrast, and not the density or velocity contrasts, that gives rise to reection energy.
In the foregoing discussion, a normal-incident compressional plane wave was considered. If the same compressional plane wave was incident at an oblique angle
to the interface, then derivation of the reected and

transmitted wave functions gets complicated, and we


nd that the reection coecient changes with angle of
incidence. Moreover, at non-normal incidence, the incident compressional-wave energy is partitioned at the
interface into four components: reected compressional,
reected shear, transmitted compressional, and transmitted shear waves.
For simplicity, consider a 2-D compressional plane
wave that impinges on a layer boundary with an angle of
incidence 0 as depicted in Figure 11.2-3a. The incident
wavefront is denoted by AC. Point A at the layer boundary acts as a Huygens secondary source and generates
its own spherical wavefronts associated with compressional and shear waves propagating in both upper and
lower media with the corresponding velocities. In Figure 11.2-3a, only the reected compressional wavefront

1810

Seismic Data Analysis

FIG. 11.2-3. Reection and refraction of an incident P -wave at a layer boundary. Medium parameters: is density, is
P -wave velocity, is S-wave velocity. (a) Reected P -wave; (b) reected S-wave; (c) refracted P -wave; (d) refracted S-wave;
(e) raypaths associated with the incident P wave, and reected and refracted P - and S-waves. The radius of the circular
wavefront associated with Huygens secondary source at A on the layer boundary is CB for the reected wave, (1 /1 ) for
the reected S-wave, (2 /1 )CB for the refracted P -wave, and (2 /1 ) for the refracted S-wave. The relationship between
the angles in (e) is given by Snells law (equation 11-10).

Reservoir Geophysics
and raypath are shown. By the time the incident wavefront at C reaches the reecting interface at B, the
spherical wavefront associated with the reected compressional wave reaches D, so that AD = CB and the
tangential line DB becomes the wavefront for the reected compressional wave. Since the incident and the
reected compressional waves travel with the same velocity, the angle of incidence 0 is equal to the angle of
reection 1 .
Now consider the reected shear wave as shown in
Figure 11.2-3b. By the time the incident wavefront at C
reaches the reecting interface at B, the spherical wavefront associated with the reected shear wave reaches
D, so that AD = (1 /1 )CB, and the tangential line
DB becomes the wavefront for the reected shear wave.
The angle for the reected shear wave 1 is no longer
equal to the angle of incidence 0 .
Huygens principle can also be applied to describe
the refracted wave at the interface. Refer to Figures
11.2-3c and 11.2-3d and note that the compressional
incident plane wave at A acts as a Huygens secondary
source and generates its own compressional wavefront
that travels into the lower medium with velocity 2 and
shear wavefront that travels into the lower medium with
velocity 2 . By the time the incident wavefront at C in
Figure 11.2-3c reaches the layer boundary at B, the
spherical wavefront associated with the refracted compressional wave reaches D, so that AD = (2 /1 )CB,
and the tangential line DB becomes the wavefront for
the refracted compressional wave. Similarly, by the time
the incident wavefront at C in Figure 11.2-3d reaches
the layer boundary at B, the spherical wavefront associated with the refracted shear wave reaches D, so that
AD = (2 /1 )CB, and the tangential line DB becomes
the wavefront for the refracted shear wave.
From the geometry of reected and refracted raypaths shown in Figure 11.2-3e, Snells law of refraction
can be deducted as
sin 0
sin 1
sin 2
sin 1
sin 2
=
=
=
=
.
1
1
2
1
2
(11 10)
Note that for all four cases in Figure 11.2-3, the horizontal distance AB at the interface is the same for the
incident and reected or the refracted wave. If AC is set
to the wavenumber along the propagation path of the
incident wave, then AB is the horizontal wavenumber
which is invariant as a result of reection or refraction.
Actually, Snells law given by equation (11-10) is a direct consequence of this physical observation.
If the compressional velocity 1 of the upper layer
is less than the compressional velocity 2 of the lower
layer, then there exists an angle of incidence such that
no refracted compressional energy is transmitted into
the lower layer. Instead, the refracted energy travels

1811

along the interface and is refracted back to the upper


layer with an angle equal to the angle of incidence. This
angle is called the critical angle of incidence for compressional waves and is given by
1
sin c =
.
(11 11a)
2
The critically refracted wave is often called the head
wave and is the basis for refraction statics (Section 3.4).
If the compressional velocity 1 of the upper layer
is less than the shear velocity 2 of the lower layer, then
there exists an angle of incidence such that no refracted
shear energy is transmitted into the lower layer. Again,
the refracted energy travels along the interface and is
refracted back to the upper layer with an angle which is
called the critical angle of P -to-S conversion given by
1
.
(11 11b)
sin c =
2
For the general case of non-normal incidence,
boundary conditions at the interface involve not only
principal stress and strain components but also the
shear stress and strain components. Again, by using the
requirement that the stress and displacement must be
continuous at the interface, a set of equations can be
derived to compute the amplitudes of the reected and
refracted P - and S-wave components associated with
an incident compressional source (Section L.5):
1
cos 1 A1 +
sin 1 B1
1
1
1
+
cos 2 A2
sin 2 B2 = cos 1 ,
2
2
(11 12a)
1
cos 1 B1
sin 1 A1 +
1
1
1
+
sin 2 A2 +
cos 2 B2 = sin 1 ,
2
2
(11 12b)
cos 21 A1 sin 21 B1
2
2
cos 22 A2
sin 22 B2 = cos 21 ,
+
1
1
(11 12c)
and
12
cos 21 B1
12
2 22 12
2 12
+
sin
2
A
+
cos 22 B2 = sin 21 .
2
2
1 12 22
1 12
(11 12d)

sin 21 A1

These are the Zoeppritz equations which can be solved


for the four unknowns, the reected compressional-wave
amplitude A1 , the reected shear-wave amplitude B1 ,
the refracted compressional-wave amplitude A2 , and

1812

Seismic Data Analysis

FIG. 11.2-4. Framework for derivation of the Zoeppritz equations. For details see Section L.5.

the refracted shear-wave amplitude B2 . Equations (1112a,b,c,d) have been normalized by the incident-wave
amplitude A0 = 1 (Section L.5).
From Snells law (equation 11-10), given the incident angle for the compressional wave and specifying
the compressional- and shear-wave velocities, the angles of reection and refraction can be computed. Substitution into equation (11-12) yields the required wave
amplitudes. These wave amplitudes, of course, depend
on the angle of incidence (Figure 11.2-3e).
Figure 11.2-4 outlines the framework for deriving
the Zoeppritz equations based on an earth model that
comprises two layers separated by a horizontal interface.
Details are left to Section L.5. Starting with the equations of motion and Hookes law, derive the wave equation for elastic waves in isotropic media in which elastic
properties are invariant in any spatial direction at any
given location. Then, use the equations of continuity
which state that the vertical and tangential stress and
stress components coincide at layer boundary, planewave solutions to the wave equation and Snells law that
relates propagation angles to wave velocities to derive

the equations for computing the amplitudes of the reected and transmitted P - and S-waves.
Refer to Figure 11.2-5 for a specic case of the
partition of energy of an incident compressional-wave
amplitude into four components. Note the signicant
changes at critical angles of refraction for the compressional and shear wave. It is important to keep in mind
that the shape of these curves varies greatly with different situations of medium parameters. Also, note that
at normal incidence no P -to-S conversion takes place,
and equations (11-12a,b,c,d) reduce to the special case
described by equations (11-5) and (11-8).
From a practical standpoint, the angle-dependency
of reection coecients implies that the reection amplitude associated with a reecting boundary varies
with source-receiver separation as well as the depth of
the reector. For suciently deep reectors and the typical source-receiver separations used in practice, the amplitude for the reected compressional wave is nearly
constant or slowly varying with oset (left of the critical angle on the curve corresponding to the reected
compressional-wave energy in Figure 11.2-5). It is this

Reservoir Geophysics

1813

FIG. 11.2-5. Partitioning of a unit-amplitude incident P -wave energy into four components reected and refracted P and S-waves (Richards, 1961.).

slowly varying portion of the P -to-P reection curve


that we have to detect from CMP data with limited
oset range and in the presence of noise. Note also from
Figure 11.2-5 that the largest P -to-S conversion occurs
beyond the critical angle, corresponding to large sourcereceiver separations.
Reection amplitude variations with angle of incidence, and therefore with oset, can be modeled using the Zoeppritz equations. For modeling the reection
amplitudes, you need well-log curves for the P -wave velocity, S-wave velocity, and density. Then, using equation (11-13a) compute the P -to-P reection amplitudes
as a function oset. Figure 11.2-6 shows a modeled CMP
gather using well data. A sonic log was rst converted
to a blocky form to simplify the modeled amplitudes on
the gather. Then, using the Gardner relation (equation

11-34a), a density prole was derived. Also, using a ratio of P -wave to S-wave velocity, a shear-wave velocity
prole was generated. The objective in this modeling
exercise was to see the eect of a change in Poissons
ratio at some depth on the amplitude variation with
oset. The Poissons ratio prole shows a change at approximately 650 ms at which time we observe a marked
variation of amplitudes with oset on the CMP gather.

Reector Curvature
Dene the ratio CE of the reection amplitude at normal incidence from a curved boundary to the reection amplitude from a at boundary at the same depth

1814

Seismic Data Analysis

FIG. 11.2-6. An example of modeling of a CMP gather based on Zoeppritz equations. (Courtesy Hampson-Russell.)

(Hilterman, 1975) as
CE =

1
,
1 + A1 z

(11 13a)

where z is the depth to the reector, and A is the reector curvature negative for synclines and positive
for anticlines. Note that for a at reector, whatever its
depth, CE = 1. But for a curved reector, the reection
amplitude at normal incidence to a synclinal interface
is greater than the case of a at reector, and the reection amplitude at normal incidence to an anticlinal
interface is smaller than the case of a at reector. The
physical basis of equation (11-13a) is that a synclinal interface focuses the energy associated with the reecting
wave, whereas an anticlinal interface defocuses it.
The eect of reector curvature on amplitude variation with oset is quantied as (Shuey et al., 1984)
CE() =

1
1+

A1 z/ cos2

(11 13b)

where is the angle of incidence. Note that equation


(11-13b) reduces to equation (11-13a) for the case of
normal incidence ( = 0). Both equations are for the
case of a 2-D reecting interface. Equation (11-13b) was
extended by Bernitsas (1990) to the case of a 3-D reecting interface with curvature in both the inline and
crossline directions.
To understand the eect of reector curvature on
amplitude variation with oset, it is convenient to study
the ratio CE()/CE( = 0) as a function of angle of
incidence (Shuey et al., 1984; Castagna, 1993). Figure
11.2-7 shows the behavior of this ratio for the cases of a
syncline and an anticlne with varying degrees of curvature. Note that, for an anticlinal interface, the curvature
eect dened by the ratio CE()/CE( = 0) decreases
with angle of incidence or oset (Figure 11.2-7c). In
practice, this means that the reection amplitudes from
an anticlinal interface at far osets are lower than those
from a at interface. Also, note that the larger the curvature dened by the ratio z/x denoted in Figure 11.2-7,

Reservoir Geophysics

1815

FIG. 11.2-7. Eect of reector curvature on angular dependence of reected wave amplitudes. The incident wave is of
compressional type. (a) A synclinal reector (x < 0), (b) an anticlinal reector (x > 0), (c) eect of curvature on the angledependent reection amplitudes associated with the reector as in (a), and (d) the same as in (c) for the reector as in (b).
(Adapted from Shuey et al., 1984.)

1816

Seismic Data Analysis

the more prominent the curvature eect. For a synclinal


interface with a mild curvature (0 > z/x > 1), the
curvature eect on amplitudes increases with angle of
incidence, and it decreases with angle of incidence for a
tight syncline (z/x < 1). Again, note that the larger
the curvature, the more prominent the curvature eect.
It is clear from the model studies summarized
above that reection amplitudes are inuenced by the
reector curvature. This is true for both normal incidence (equation 11-13a) and non-normal incidence
(equation 11-13b). Reection amplitudes must be corrected for the reector curvature by prestack time migration before prestack amplitude inversion to derive
the AVO attributes which are discussed in this section.
Similarly, reection amplitudes must be corrected for
the reector curvature by poststack time migration before poststack amplitude inversion to derive the acoustic impedance attribute which is discussed in the next
section.

AVO Equations
Consider the two elastic half-space layers in Figure 11.23e. The Zoeppritz equations (11-14) can be solved for
the reected and refracted P - and S-wave amplitudes,
A1 , B1 , A2 , and B2 . However, our interest in exploration
seismology is largely the angle-dependency of the P -toP reections given by the coecient A1 . Specically,
we wish to infer or possibly estimate elastic parameters
of reservoir rocks from reection amplitudes and relate
these parameters to reservoir uids.
The exact expression for A1 derived from the solution of the Zoeppritz equations (11-13) is complicated
and not intuitive in terms of its practical use for inferring petrophysical properties of reservoir rocks. The
rst approximation to the Zoeppritz equation for P -toP reection amplitude is given by Bortfeld (1961) as
R(1 ) =

1
2 2 cos 2
ln
2
1 1 cos 1
+ 2+

ln
ln

2
1

2
2 1
ln
1
1 2

12 22
sin2 1 .
12
(11 14)

The arrangement of the two terms on the right-hand


side of equation (11-14) is based on separating the
acoustic (the rst term) and the elastic (the second
term) eects on reection amplitudes. As such, equation (11-14) does not explicitly indicate angle- or osetdependence of reection amplitudes; therefore, its practical implementation for AVO analysis has not been considered.

Instead, we shall use the approximation provided


by Aki and Richards (1980) as the starting point for deriving a series of practical AVO equations. Now that we
only need to deal with the P -to-P reection amplitude
A1 , we shall switch to the conventional notation by replacing A1 with R() as the angle-dependent reection
amplitude for AVO analysis.
By assuming that changes in elastic properties of
rocks across the layer boundary are small and propagation angles are within the subscritical range, the exact
expression for R() given by the Zoeppritz equation can
be approximated by (Aki and Richards, 1980)
1
1 + tan2
2

R() =
+

2
4 2 sin2

2
1
1 4 2 sin2
2

(11 15)

where = (1 + 2 )/2, average P -wave velocity and


= (2 1 ), = (1 + 2 )/2, average S-wave velocity and = 2 1 , = (1 + 2 )/2, average
density and = 2 1 , and = (1 + 2 )/2, average
of the incidence and transmission angles for the P -wave
(Figure 11.0-2e).
Figure 11.2-8 shows the angle-dependent reection
amplitude associated with an interface with contrast in
P - and S-wave velocities and densities based on the exact Zoeppritz equation, the Bortfeld approximation described by equation (11-14), and the Aki-Richards approximation described by equation (11-15) (Smith and
Gidlow, 1987). Note that these approximations closely
follow the exact solution within the range of angles of incidence that are achievable by the recording of seismic
data used in exploration. For very shallow reectors,
however, the approximate solutions would deviate from
the exact solution signicantly at very wide angles of
incidence.
Note that the Aki-Richards approximation to the
angle-dependent reection amplitude R() given by
equation (11-15) has three parts in terms of /
which describes the fractional change in P -wave velocity across the layer boundary and hence may be referred
to as the P -wave reectivity, / which describes the
fractional change in the S-wave velocity across the layer
boundary and hence may be referred to as the S-wave
reectivity, and / which describes the fractional
change in density across the layer boundary.
In practice, we do not observe the separate eects
of P -wave reectivity /, S-wave reectivity /
and fractional change in density / on the reection
amplitudes R(). Instead, we observe changes in reection amplitudes as a function of angle of incidence. In
fact, it is the elastic parameters such as the P -wave

Reservoir Geophysics

1817

FIG. 11.2-8. P -to-P reection amplitude as a function of angle of incidence computed by using the exact Zoeppritz equation,
and Bortfeld (equation 11-14) and Aki-Richards (equation 11-15) approximations (Smith and Gidlow, 1987).

1818

Seismic Data Analysis

reectivity /, S-wave reectivity / and fractional chnage in density / that we wish to estimate from the observed angle-dependent reection amplitudes. To use the Aki-Richards equation (11-15) in
the inversion of reection amplitudes for these elastic parameters, we rst need to recast it in successive
ranges of angle of incidence. This change of philosophy
in arranging the terms in the Aki-Richards equation
(11-15) was rst introduced by Shuey (1985) and led
to practical developments in AVO analysis. The new
arrangement in terms of successive ranges of angle of
incidence is given by
R() =

1
2

1
2
2
4 2
2 2
sin2
2

1
2

tan2 sin2 .

1
2

1 2
2
1

(11 17a)

to perform the necessary dierentiation

2 (1 )(1 2)

(11 17b)

The compressional-wave velocity is given by equation

(11 17c)

where is Lames constant.


We also dene the P -wave reection amplitude RP
at normal incidence as
RP =

1
2


.
+

(11 18)

Substitute equations (11-17a), (11-17b), and (11-18)


into the Aki-Richards equation (11-16) and perform
some algebraic simplication to obtain
R() = RP

(11 16)

Another practical matter of concern is that the


Aki-Richards equation (11-15) or any of its modications that we shall derive in this section describe the
modeled reection amplitudes as a function of angle of
incidence. However, the observed reection amplitudes
are available from CMP data as a function of oset. A
need then arises either to transform the model equation for the reection amplitudes from angle to oset
coordinates (Demirbag and Coruh, 1988) or to actually
transform the CMP data from oset to angle coordinates. While the rst approach is theoretically appealing, the practical schemes are based on the latter approach. We have already discussed such a transformation in Section 6.4 the Radon transform using the
linear moveout equation or its robust variation in the
form of slant stacking. Figure 11.2-9 shows Zoeppritz
amplitude curves as a function of angle of incidence and
oset (Demirbag and Coruh, 1988).
Based on the theoretical conjecture made earlier by
Koefoed (1955) that the elastic property that is most
directly related to angular dependence of reection coecient R() is Poissons ratio , Shuey (1985) introduced a variable transformation from S-wave velocity
to . The relationship between the two variables is
given by equation (L-49) which we rewrite below as
2 =

(11-3b) and the shear-wave velocity is given by equation (L-47) which is rewritten below as


1
2
+
2

1
2

12

+
sin2
1
(1)2

tan2 sin2 .

Dene a new term H


/
,
H=
/ + /

(11 19)

(11 20)

and by way of equation (11-18) note that

(11 21a)
= 2 RP H.

Next combine equations (11-20) and (11-21a) to obtain

= 2 RP (1 H).

(11 21b)

Finally, substitute equations (11-21a) and (11-21b) into


the second term on the right-hand side of equation (1119) to obtain
R() = RP + RP H0 +
+

1
2

sin2
(1 )2

tan2 sin2 ,

(11 22)

where
H0 = H 2(1 + H)

1 2
.
1

(11 23)

Equation (11-22) is known as Shueys three-term


AVO equation. The rst term RP is the reection amplitude at normal incidence. At intermediate angles
(0 < < 30 degrees), the third term may be dropped,
thus leading to a two-term approximation
R() = RP + G sin2 ,
where

(11 24)

Reservoir Geophysics

1819

1820

Seismic Data Analysis


G = RP H0 +

.
(1 )2

(11 25)

Equation (11-24) is known as Shueys two-term


AVO equation. In practice, amplitudes picked along
a moveout-corrected event on a CMP gather plotted
against sin2 can be tted to a straight line. The slope
of the line gives the AVO gradient attribute and the ordinate at zero angle gives the AVO intercept attribute.
The AVO gradient given by equation (11-25) is directly
related to change in Poissons ratio , which in turn,
is directly related to uid saturation in reservoir rocks.
The AVO intercept attribute represents the reectivity RP at normal incidence. Therefore, the AVO intercept attribute, in lieu of conventional stack, can be
used as input to derive the acoustic impedance attribute
(Section 11.3), which is indirectly related to porosity in
reservoir rocks.
Shown in Figure 11.2-10a is a portion of a section derived from 2-D prestack time migration. The
objective is to identify uid-saturated reservoir zones
at the apex and the anks of the structural closure.
This image section is derived from the stacking of
common-reection-point (CRP) gathers associated with
the prestack time-migrated data. The CRP gather in
Figure 11.2-10b shows three events with amplitude variations with oset which are plotted in Figure 11.2-10c.
By using Shueys equation (11-24), the AVO gradient
and intercept sections are computed from the CRP
gathers as shown in Figures 11.2-11 and 11.2-12, respectively. Note that the gradient section exhibits a group
of AVO anomalies in the vicinity of the structural apex,
possibly indicating uid-saturated reservoir rocks.
At large angles of incidence beyond 30 degrees, the
third term in equation (11-22) gradually becomes dominant. Note that this term is related directly to fractional change in P -wave velocity, /. So, not only
the reection traveltimes at far osets (Section 3.1) corresponding to large angles of incidence, but also the reection amplitudes at large angles of incidence make
the biggest contribution to the resolution needed to estimate the changes in P -wave velocities.
The two-term equation (11-24) can be specialized
for a specic value of Poissons ratio, = 1/3 and H0 =
1 so that equation (11-25) takes the form
9
(11 26)
G = RP + ,
4
which can be solved for the change in Poissons ratio
across a layer boundary
4
RP + G .
(11 27)
9
This is the AVO attribute equation for estimating
changes in Poissons ratio (Hilterman, 1983). Actually,
=

as described by equation (11-27), this attribute is the


scaled sum of the AVO intercept RP and AVO gradient
G attributes.
By recasting the rst-order approximation to the
Zoeppritz equation, Wiggins et al. (1984) and Spratt
et al. (1984) derived a practical expression for S-wave
reectivity. First, drop the term with tan2 in equation
(11-16) and rearrange the remaining terms to obtain
R() =
+

1
2

1
+
2

+ 2

2 1
8 2
+
2

2 1
sin2

2
2

sin2
(11 28)

such that, much like the denition for the P -wave reectivity RP given by equation (11-18), an expression
for the S-wave reectivity RS
RS =

1
2

(11 29)

can be explicitly inserted back into equation (11-28) to


get
R() = RP +
+

RP 8

2
1

2
RS
2

sin2 .

sin2
(11 30)

Set / = 0.5 to make the last term on the righthand side vanish and obtain (Spratt et al., 1984)
R() = RP + RP 2 RS sin2 .

(11 31)

This equation is of the form given by equation (11-24)


where
G = RP 2 R S ,

(11 32)

which can be rewritten explicitly in terms of the shearwave reectivity RS


1
RP G .
(11 33)
2
This is the AVO attribute equation for estimating the
shear-wave reectivity. Given the AVO intercept RP
and AVO gradient G attributes, simply take half of
the dierence between the two attributes to derive the
shear-wave reectivity RS as described by equation (1133).
Figure 11.2-13 shows the reection amplitudes as
a function of angle as predicted by equation (11-31) for
three combinations of RP and RS . Note that equation
(11-31) is a good approximation to the P -to-P reection
RS =

Reservoir Geophysics

1821

FIG. 11.2-10. (a) A portion of a prestack time-migrated section; (b) portion of a common-reection-point gather in the
vicinity of the structural apex in (a); (c) reection amplitudes as a function of oset measured along three events indicted by
the red, horizontal trajectories in (b).

1822

Seismic Data Analysis

FIG. 11.2-11. The AVO gradient section dened by G in equation (11-24); (b) close-up of (a) in the vicinity of the structural
apex.

Reservoir Geophysics

1823

FIG. 11.2-12. (a) The AVO intercept section dened by RP in equation (11-24); (b) close-up of (a) in the vicinity of the
structural apex.

1824

Seismic Data Analysis

FIG. 11.2-13. P -to-P reection amplitude as a function of angle of incidence computed by using the exact Zoeppritz equation
(the three curves labeled as A) and the approximate form given by equation (11-31) (the three curves labeled as B) (Spratt
et al., 1984).

amplitudes as predicted by the exact Zoeppritz equation


up to nearly 30 degrees of angles of incidence.
Return to the Aki-Richards equation (11-15) and
consider the case of N -fold CMP data represented in the
domain of angle of incidence. Note that the reection
amplitude R() is a linear combination of three elastic
parameters P -wave reectivity /, S-wave reectivity / and fractional change in density /.
Smith and Gidlow (1987) argue in favor of solving
for only two of the three parameters by making use of
the empirical relation between density and P -wave
velocity (Gardner et al., 1974):
= k1/4 ,

(11 34a)

where k is a scalar. This relation holds for most watersaturated rocks. Dierentiate to get
1

=
.

(11 34b)

Now substitute equation (11-34b) into the original


Aki-Richards equation (11-15) and combine the terms
with /
R() =

1
2
1
1 + tan2 +
1 4 2 sin2
2
8

sin2
.
2

(11 35)

Simplify the algebra to obtain the desired two-parameter model equation for prestack amplitude inversion
(Section L.6)
1

5 1 2
2
2

sin
+
tan

4
sin2
.
8 2 2
2

(11 36)
Redene the coecients ai and bi and write the
discrete form of equation (11-36)

R()=

Ri = ai

+ bi
,

(11 37)

where
ai =

5 1 2
1

sin2 i + tan2 i
2
8 2
2

(11 38a)

and
2
sin2 i .
(11 38b)
2
Consider the case of N -fold CMP data represented
in the domain of angle of incidence. We have, for each
CMP location and for a specic reection event associated with a layer boundary, N equations of the form
as equation (11-37) and two unknowns / and
/. We have more equations than unknowns; hence,
we encounter yet another example of a generalized linear inversion (GLI) problem (Section J.1). The objective is to determine the two parameters such that the
bi = 4

Reservoir Geophysics

1825

FIG. 11.2-14. (a) P -wave and (b) S-wave reectivity sections derived from the Smith-Gidlow inversion of prestack amplitudes
(equation 11-39).

1826

Seismic Data Analysis

FIG. 11.2-15. Laboratory measurements of (a) P -wave and (b) S-wave reectivities at a gas-brine interface in sandstones
(Spratt et al., 1984). The horizontal axis in both plots denotes induration which is a measure of rock strength.

FIG. 11.2-16. Portions of (a) P -wave and (b) an S-wave stacked section that show anomalies associated with lithology (coal)
and uid content (gas) (Spratt et al., 1984).

Reservoir Geophysics

1827

FIG. 11.2-17. A sketch of crossplot of P - to S-wave velocity (Castagna et al., 1985; Fatti et al., 1994).

dierence between the modeled reection amplitudes Ri


represented by equation (11-37) and the observed reection amplitudes Xi is minimum in the least-squares
sense (Smith and Gidlow, 1987).
The GLI solution in matrix form is given by (Section L.6)

N
N 2
N
i ai
i ai bi
i ai Xi

.
=
N
N 2
N
i bi
i ai bi
i bi X i

(11 39)
This matrix equation is solved for the two parameters P -wave reectivity / and S-wave reectivity /. Note from the denitions of the coecients
ai and bi given by equations (11-38a,b) that you have
to choose a value for the ratio / to compute the two
reectivities.
Shown in Figure 11.2-14 are the P - and S-wave reectivity sections that were derived from prestack amplitude inversion applied to the CRP gathers associated

with the data shown in Figure 11.2-10a. Note the AVO


anomalies along the anks of the structural closure. It is
apparent that some of the AVO anomalies in the P -wave
reectivity section stand out more distinctively as compared with those in the S-wave reectivity section. This
may be related to the fact that changing from brine to
gas causes a small change in S-wave reectivity, while it
causes a signicant change in P -wave reectivity. Such
inference is supported by the laboratory measurements
of P - and S-wave reectivities at a gas-brine interface
in sandstones (Spratt et al., 1984). Figure 11.2-15 shows
P -wave reectivities for a group of reservoir sandstones
in which the pore uid changes from brine to gas. For
rocks with low induration (weak rocks), there is a large
change in P -wave reectivity, while for rocks with high
induration the change becomes less signicant (Figure
11.2-15a). The change in S-wave reectivity is negligible for rocks with low and high induration (Figure
11.2-15b). A demonstrative eld data example of P - and
S-wave reectivity contrasts that arise from a change in

1828

Seismic Data Analysis

FIG. 11.2-18. Crossplot of P - to S-wave velocity measured down a borehole (Fatti et al., 1994).

lithology and uid saturation is shown in Figure 11.216.


Refer to equation (11-17b) and note that the difference between the P -wave and S-wave reectivities is
related to change in Poissons ratio a direct hydrocarbon indicator. In fact, Smith and Gidlow (1987)
have coined the term pseudo-Poisson reectivity to describe the dierence between the P -wave and S-wave
reectivities

.
(11 40)

Note from equation (11-17b) that the pseudoPoisson reectivity


/
is not exactly the same as
what may be referred to as the proper Poisson reectivity /. If we dene the ratio

=
(11 41a)

by dierentiation, we can derive the dierence relation


given by equation (11-40) and thus show that
(/)

=
.

(11 41b)

Castagna et al. (1985) dened a straight line in


the plane of S-wave velocity versus P -wave velocity as
shown in Figure 11.2-17. This is called the mudrock line

and is represented by the equation


= c0 + c1 ,

(11 42)

where the scalar coecients c0 and c1 are empirically


determined for various types of rocks. Suggested values
for these scalars are c0 = 1360 and c1 = 1.16 for watersaturated clastics (Castagna et al., 1985). Gas-bearing
sandstones lie above the mudrock line, and carbonates
lie below the mudrock line as shown on a crossplot of P and S-wave velocities sketched in Figure 11.2-17. The
crossplot shown in Figure 11.2-18 is based on log measurements at a gas-producing well (Fatti et al., 1994).
Note the separation of gas-sandstone cluster from the
water-sandstone and shale clusters in the manner as
sketched in Figure 11.2-17.
A way to quantify the prospectivity of the reservoir
rock of interest is by dening a uid factor attribute that
indicates the position of the rock property with respect
to the mudrock line. First, apply dierentiation to both
sides of equation (11-42) and note that


= c1
.

(11 43)

Then, dene the uid factor F


F =


c1
.

(11 44)

Reservoir Geophysics

1829

FIG. 11.2-19. (a) Pseudo-Poisson and (b) uid factor attribute sections derived by using equations (11-40) and (11-44),
respectively.

1830

Seismic Data Analysis

FIG. 11.2-20. (a) The and (b) attribute sections derived by using equations (11-52b) and (11-52c), respectively.

Reservoir Geophysics

1831

FIG. 11.2-21. (a) Crossplot of P -wave impedance versus S-wave impedance, and (b) crossplot of versus AVO attributes.

If F is close to zero, it means that you have watersaturated rock. If you have a gas-saturated sandstone,
F will be negative at the top and positive at the base
of the reservoir unit (Smith and Gidlow, 1987).
Figure 11.2-19 shows the pseudo-Poisson and uidfactor sections that were derived from prestack amplitude inversion applied to the CRP gathers associated
with the data shown in Figure 11.2-10a. Note the distinctive AVO anomalies along the anks of the structural closure.
The two parameters / and /, estimated
by using the least-squares solution given by equation
(11-39), represent fractional changes in P - and S-wave
velocities. As such, they are related to P - and S-wave
reectivities, IP /IP and IS /IS , respectively, where
IP and IS are the P - and S-wave impedances given by
IP =

(11 45a)

and
IS = .

(11 45b)

From Section L.6, the P - and S-wave reectivities


are given by

and

IP

=
+
IP

(11 46a)

IS

=
+
.
IS

(11 46b)

Assuming that the density obeys Gardners relation


given by equation (11-34a), the P -wave reectivity
IP /IP is simply the fractional change of the P -wave
velocity / scaled by a constant, whereas the S-wave
reectivity IS /IS is a linear combination of the fractional changes of the P - and S-wave velocities, /
and /, respectively.

1832

Seismic Data Analysis

Reservoir Geophysics
A direct estimation of the P - and S-wave reectivities given by equations (L-46a,b) can be made by using
an alternative formulation of prestack amplitude inversion which is based on transforming the Aki-Richards
equation (11-15) to the new variables IP /IP and
IS /IS (Goodway et al., 1998). Solve equation (11-46a)
for / and equation (11-46b) for /, and substitute into the Aki-Richards equation (11-15). Following
the algebraic simplication, we obtain (Section L.6)
R() =

1
1 + tan2
2

IP
IS
2
4 2 sin2
IP

IS

1
2

tan2 2 2 sin2
.
2

(11 47)

Goodway et al. (1998) have implemented a specic


form of equation (11-47) to derive the AVO attributes
IP /IP and IS /IS . For a specic value of / = 2
and small angles of incidence for which tan sin ,
the third term in equation (11-47) vanishes. Compare
equations (11-18) and (11-46a), and equations (11-29)
and (11-46b), and note that equation (11-47) with the
remaining terms takes the form
R() = 1 + tan2 RP 2 sin2 RS .

(11 48)

The resulting equation then is solved for the P - and


S-wave reectivities, IP /IP = 2RP and IS /IS =
2RS , respectively. Again, consider the case of N -fold
CMP data represented in the domain of angle of incidence. In discrete form, equation (11-48) can be rewritten as
Ri = ai RP + bi RS ,

(11 49)

where i is the trace index and the coecients ai and bi


are given by
ai =

1
1 + tan2 i
2

(11 50a)

and
bi = 2 sin2 i .

(11 50b)

The reection amplitude Ri in equation (11-49) is a linear combination of the two parameters, RP and RS . As
for the Smith-Gidlow equation (11-37), the two parameters, RP and RS , can be estimated for a specic event
from CMP data using the least-squares solution given
by

N 2
N
RP
i ai
i ai bi
i ai Xi

=
.
N
N 2
N
RS
i bi
i ai bi
i bi X i
(11 51)
Following the estimation of the P -wave reectivity RP and the S-wave reectivity RS using the leastsquares solution given by equation (11-51), the P -wave

1833

impedance IP and the S-wave impedance IS can be


computed by integration.
By using the impedance attributes IP and IS ,
Goodway et al. (1998) compute two additional AVO attributes in terms of Lames constants scaled by density
and . Substitute equation (11-3b) into (11-45a)
to get the relation
+ 2 = IP2 ,

(11 52a)

and substitute equation (11-17c) into equation (11-45b)


to get the relation
= IS2 .

(11 52b)

Note from equations (11-52a,b) that


= IP2 2 IS2 .

(11 52c)

Figure 11.2-20 shows the and AVO attribute


sections which were derived from prestack amplitude
inversion applied to the CRP gathers associated with
the data shown in Figure 11.2-10a. As in the case of
the pseudo-Poisson and uid-factor AVO attribute sections shown in Figure 11.2-19, note the distinctive AVO
anomalies along the anks of the structural closure.
Goodway et al. (1998) convincingly demonstrates that
separation of gas-bearing sands from tight sands and
carbonates is much better with the crossplot of the
Lame attributes (, ) in contrast with the crossplot
of the P - and S-wave reectivities or impedances (Figure 11.2-21).
Figure 11.2-22 outlines a summary of the AVO
equations that are described in this section based on the
various approximations to the Aki-Richards equation
(11-15). Start with Shueys arrangement of the terms
in the Aki-Richards equation, which itself is an approximation to the exact expression for the P -to-P reection
amplitudes given by the Zoeppritz equations. Then, apply a transformation from S-wave velocity to Poissons
ratio, and drop the third term to get a simple expression for the P -to-P reection amplitude that is a linear
function of sin2 . This linear approximation yields the
AVO intercept and gradient attributes.
Alternatively, you may drop the third term in the
original Shuey arrangement given by equation (11-16)
and apply Wiggins rearrangement of the terms. Then,
assume the specic case of the P -wave velocity to be
twice the S-wave velocity to obtain the equation that
yields, once again, the AVO intercept and gradient attributes. The Wiggins AVO intercept and gradient attributes, unlike the Shuey counterparts, directly lead to
deriving the P -wave and S-wave reectivity attributes,
albeit only under the assumption that the P -wave velocity is twice the S-wave velocity.

1834

Seismic Data Analysis

FIG. 11.2-23. A raw shot record from the AVO line presented in Section 11.2. (Data courtesy Britannia Gas Ltd.)

Reservoir Geophysics

1835

FIG. 11.2-24. The shot record in Figure 11.2-23 after muting for guided waves and t2 -scaling for geometric spreading
correction. (Data courtesy Britannia Gas Ltd.)

1836

Seismic Data Analysis

FIG. 11.2-25. The shot record in Figure 11.2-24 after spiking deconvolution. (Data courtesy Britannia Gas Ltd.)

Reservoir Geophysics

1837

1838

Seismic Data Analysis

FIG. 11.2-27. Autocorrelogram (a) of the shot record in Figure 11.2-23, (b) after muting for guided waves and t2 -scaling for
geometric spreading correction, and (c) after spiking deconvolution.

Reservoir Geophysics
Returning to the three-term Aki-Richards equation (11-15), you may reduce it to the two-term SmithGidlow equation (11-36) by using Gardners relation of
density and P -wave velocity. Then, for a specied ratio
of P - to S-wave velocity, perform inversion of prestack
amplitudes associated with target events on CMP or
CRP gathers to obtain the P - and S-wave reectivity
attributes. In addition, using these attributes to derive
two more AVO attributes psuedo-Poisson and uid
factor.
Finally, you may recast the Aki-Richards equation
(11-15) in terms of P - and S-wave reectivities (equations 11-46a,b) and assume that / = 2 to obtain the
two-term Goodway et al. equation (11-48). Again, perform inversion of prestack amplitudes associated with
target events on CMP or CRP gathers to obtain the P and S-wave impedance attributes. Subsequently, by using the relations between impedances and Lames constants, you can compute two additional AVO attributes
and .
The AVO equations derived in this section are all
expressed in terms of angle of incidence (equations 1124, 11-31, and 11-36). In practice, the mapping of amplitudes associated with a reection event on a CMP
gather from oset to angle of incidence needs to be performed. Assume that the CMP gather is associated with
a horizontally layered earth model so that the event
moveout on the CMP gather is given by the hyperbolic
equation
t2 = t20 +

x2
2
vrms

(11 53)

where t is the two-way traveltime from the source to


the at reecting interface back to the receiver, t0 is
the two-way zero-oset time, x is the oset, and vrms
is the rms velocity down to the reector. The ray parameter p is given by the stepout dt/dx measured along
the hyperbolic moveout trajectory (Section 6.3). Apply
dierentiation to equation (11-53) to get
1 x
dt
= 2
.
dx
vrms t

(11 54)

Since the ray parameter p is also expressed as the ratio


of sin /vint , where vint is the interval velocity above
the reecting interface, it follows that
vint x
.
(11 55)
sin = 2
vrms t
By using equation (11-55), reection amplitudes on
a CMP gather can be transformed from oset to angle
domain, and subsequently used in the AVO equations
(11-24), (11-31), (11-36), and (11-48).

1839

Processing Sequence for AVO Analysis


There are three important aspects of a processing sequence tailored for AVO analysis.
(a) The relative amplitudes of the seismic data must be
preserved throughout the analysis so as to recognize amplitude variation with oset. This requirement often leads to an application of a parsimonious sequence of signal processing to avoid distortion of amplitudes by undesirable eects of some
processing algorithms.
(b) The processing sequence must retain the broadest
possible signal band in the data with a at spectrum within the passband.
(c) Prestack amplitude inversion to derive the AVO
attributes must be applied to common-reectionpoint (CRP) gathers (Section 5.4), not to commonmidpoint (CMP) gathers. This is because all the
AVO equations described in this section are based
on a locally at earth model that can be related to
CRP raypaths but not to CMP raypaths. Specically, the CRP gathers are associated with events in
their migrated positions, whereas the CMP gathers
are associated with events in their unmigrated positions. The AVO equations described in this section
are all based on a horizontally layered earth model
that can be related to CRP raypaths but not to
CMP raypaths. Prestack time migration compensates for the eect of reector curvature (equation
11-13b) on amplitude variation with oset so that
the reection amplitudes in each of the resulting
CRP gathers can be associated with a locally at
earth model.
We shall consider a marine 2-D seismic data set
from the North Sea recorded over a producing gas eld.
Data specications for the line are listed in Table 11-3.
The prestack signal processing that complies with
the above requirements includes the following steps:
(a) If a reliable source signature is available, perform
signature processing (Section 2.5) to convert the
source waveform to its minimum-phase equivalent.
In the present study, no source signature was available. Apply a mute to shot records to remove
the guided waves associated with the dispersive,
normal-mode propagation within the water layer
(Section F.1).
(b) Apply t2 -scaling to correct for geometric spreading.
A gain function for geometric spreading correction
that depends on primary velocities will overcorrect
(text continues on p. 1848)

1840

Seismic Data Analysis

Reservoir Geophysics

1841

FIG. 11.2-29. The CRP gather at well location CRP 1134. The reservoir zone coincides with the time window indicated by
the area within the rectangle.

1842

Seismic Data Analysis

FIG. 11.2-30. The unmigrated stack associated with the prestack time-migrated data as in Figure 11.2-28.

Reservoir Geophysics

FIG. 11.2-31. The same stack as in Figure 11.2-30 after poststack deconvolution.

1843

1844

Seismic Data Analysis

FIG. 11.2-32. Migration of the stack in Figure 11.2-31 using the rms velocity eld derived from the velocity analysis of the
CRP gathers as in Figure 11.2-28. The area within the rectangle corresponds approximately to the sections in Figures 11.2-41
and 11.2-42.

Reservoir Geophysics

1845

1846

Seismic Data Analysis

FIG. 11.2-34. Autocorrelogram (a) of the stack in Figure 11.2-30, (b) of the stack in Figure 11.2-31, and (c) of the stack in
Figure 11.2-32.

Reservoir Geophysics

1847

1848

Seismic Data Analysis

Table 11-3. Data specications for the line used in


AVO analysis.
Line Length
Shot Spacing
Receiver Spacing
CMP Spacing
Minimum Oset
Maximum Oset
Number of Receivers
Fold of Coverage
Sampling interval

23.5
37.5
18.75
9.375
150
3050
156
78
2

km
m
m
m
m
m
ms

the amplitudes of multiples. Figure 11.2-23 shows


a recorded shot record and 11.2-24 shows the same
record after guided-wave muting geometric spreading correction.
(c) Apply time-invariant spiking deconvolution with
a suciently long operator length (Figure 11.225). For land data, it may be necessary to apply surface-consistent amplitude corrections and
surface-consistent deconvolution to account for the
near-surface eects on amplitudes (Section B.8).
(d) When it is required, apply time-variant spectral
whitening to account for the nonstationarity of the
waveform.
Figure 11.2-26 shows the amplitude spectrum of the
shot record in Figure 11.2-25 computed after each of the
steps in the sequence prescribed above. To begin with,
note that much of the energy in the raw shot record is
conned to the shallow portion and is mostly associated
with the guided waves. The t2 -scaling restores the amplitudes in the deeper portion of the record. Spiking deconvolution followed by a wide-bandpass lter gives the
desired shape to the spectrum with a at, broadband
character. Figure 11.2-27 shows the autocorrelograms
of the shot record in Figure 11.2-25, again, computed
after each of the steps in the sequence prescribed above.
Note that the autocorrelogram of the raw shot record
is dominated by the guided-wave energy at far osets
on the left and the energy associated with short period
multiples and reverberations at near osets on the right.
The guided-wave muting followed by t2 -scaling exposes
the energy associated with multiples that range from
a short- to a moderate-period. Note that spiking deconvolution with an operator length of 480 ms (about
one-tenth of the total length of the shot record) has
removed a signicant amount of the multiple energy
from the record. Both the amplitude spectrum (Figure
11.2-26c) and the autocorrelogram (Figure 11.2-27c) of
the deconvolved shot record indicate that time-variant

spectral whitening does not have to be included in the


sequence.
The prestack data after signal processing are ready
now for prestack time migration to generate the CRP
gathers (Section 5.3).
(a) Sort the signal-processed shot records to CMP
gathers and perform velocity analysis at sparse intervals along the line. To increase the accuracy of
prestack amplitude inversion and thus increase the
condence level for the AVO attributes, the fold of
coverage and the oset range of the recorded data
must not be reduced during the processing for any
reason.
(b) Apply NMO correction using the at-event velocities. In this case, only three velocity functions were
used in creating the velocity eld for NMO correction.
(c) When required, attenuate those multiples that have
not been handled by deconvolution during the signal processing stage described above. Data suitable
for AVO analysis are often associated with nearhorizontal earth models or low-relief structures.
Hence, techniques based on the discrete Radon
transform (Section 6.4) often are best suited for
amplitude-preserving multiple attenuation.
(d) Sort the data to common-oset sections and apply
DMO correction.
(e) Perform zero-oset, frequency-wavenumber timemigration on each of the common-oset sections
after NMO and DMO corections using a single, vertically varying velocity function. Again, data suitable for AVO analysis are generally associated with
near-horizontal earth models or low-relief structures. Hence, use of a single, vertically varying velocity function and a zero-oset phase-shift algorithm for common-oset migration often is appropriate.
(f) Sort the data to CRP gathers and apply inverse
NMO correction using the velocity eld from step
(b).
(g) Perform velocity analysis at frequent intervals
along the line, and create a velocity eld associated with the migrated data.
(h) Apply NMO correction using the velocity eld from
step (g) and, when required, as it was in the present
case study, perform residual moveout velocity analysis to remove any residual moveout errors that
may be present in the CRP gathers. The residual moveout analysis actually is a good practice
whatever the degree of residual moveout errors in
the data, since prestack amplitude inversion is extremely sensitive to how at the events are in the
CRP gathers. Figure 11.2-28 shows selected CRP

Reservoir Geophysics

1849

FIG. 11.2-36. Log data measured at well location CRP 1134: (a) the sonic log, (b) the density log, (c) synthetic seismogram
using a 90-degree phase-rotated band-limited wavelet (plotted at time 3.5 s), and (d) the reectivity series computed from
(a) and (b) and used in creating (c).

1850

Seismic Data Analysis

Reservoir Geophysics

1851

FIG. 11.2-38. (b) A close-up portion of the CRP gather 1134 as in Figure 11.2-37a at the well location and at reservoir
level, (b) the corresponding modeled CRP gather as in Figure 11.2-37b.

gathers that exhibit mostly at events. A close-up view


of the CRP gather (Figure 11.2-29) at the well location
(CRP 1134) shows the reservoir zone highlighted by the
rectangle situated between 3.1-3.3 s.
(i) Stack the CRP gathers to obtain the image section from prestack time migration. Multiples that
may have remained in the data will exhibit some
residual moveout that can be exploited to further
attenuate them during stacking.
(j) For completeness of the prestack time migration
sequence, unmigrate the CRP stack from step (i)
using the same vertically varying velocity function
as in step (e) and a phase-shift zero-oset waveeld
modeling algorithm (Figure 11.2-30).
(k) Next, apply poststack spiking deconvolution, often
using the same parameters as for prestack deconvolution (Figure 11.2-31).
(l) Remigrate the modeled zero-oset section from
step (k) using the migration velocity eld derived
from the CRP gathers in step (g) (Figure 11.2-32).
In this instance, any lateral velocity variations implied by the migration velocity eld need to be accounted for by the migration algorithm of choice.
Poststack amplitude inversion to derive the acoustic impedance attribute (Section 11.3) may then be

applied to the nal product from prestack time migration (Figure 11.2-32). Figure 11.2-33 shows the
amplitude spectrum of the CRP stack after each
step of the poststack processing. The amplitude
spectrum of the CRP stack from step (i) shows the
attenuation of high frequencies (Figure 11.2-33a),
which are restored by deconvolution (Figure 11.233b), but are slightly shifted to lower frequencies by
migration (Figure 11.2-33c). The autocorrelogram
of the CRP stack from step (i) exhibits remnants
of the water-bottom multiples (Figure 11.2-34a),
which are largely attenuated by poststack deconvolution (Figure 11.2-34b).

Derivation of AVO Attributes by


Prestack Amplitude Inversion
Following the prestack signal processing and prestack
time migration designed for AVO analysis, prestack amplitude inversion is applied to the CRP gathers from
prestack time migration. Whatever the basis for the
AVO equations (Figure 11.2-22), derivation of the AVO
attributes requires knowledge of the angle of incidence.
(text continues on p. 1859)

1852

Seismic Data Analysis

FIG. 11.2-39. Amplitude variation with oset extracted from the CRP gather as in Figure 11.2-38 at the well location and
transformed to amplitude variation with angle by using the angle of incidence information shown in Figure 11.2-37c. See text
for details.

Reservoir Geophysics

1853

FIG. 11.2-40. The stacks of the CRP data from prestack time migration as in Figure 11.2-28. The solid triangle denotes the
well location at the surface, and the insertion below the well location is the synthetic equivalent of the sections shown. The
dotted near-vertical line to the right is the well trajectory within the time gate corresponding to the portion of the sections
shown.

1854

Seismic Data Analysis

FIG. 11.2-41. Part 1: AVO attributes derived from prestack amplitude inversion of the CRP data as in Figure 11.2-28. The
solid triangle denotes the well location at the surface, and the insertion below the well location is the corresponding synthetic
AVO attribute. The dotted near-vertical line to the right is the well trajectory within the time gate corresponding to the
portion of the sections shown.

Reservoir Geophysics

1855

FIG. 11.2-41. Part 2: AVO attributes derived from prestack amplitude inversion of the CRP data as in Figure 11.2-28. The
solid triangle denotes the well location at the surface, and the insertion below the well location is the corresponding synthetic
AVO attribute. The dotted near-vertical line to the right is the well trajectory within the time gate corresponding to the
portion of the sections shown.

1856

Seismic Data Analysis

FIG. 11.2-41. Part 3: AVO attributes derived from prestack amplitude inversion of the CRP data as in Figure 11.2-28. The
solid triangle denotes the well location at the surface, and the insertion below the well location is the corresponding synthetic
AVO attribute. The dotted near-vertical line to the right is the well trajectory within the time gate corresponding to the
portion of the sections shown.

Reservoir Geophysics

1857

FIG. 11.2-42. The postreservoir zone in the vicinity of the well location CRP 1134 (within the blue rectangle), the reservoir
zone (within the red rectangle), and the prereservoir zone (within the green rectangle) labeled on (a) the intercept and (b)
the gradient sections.

1858

Seismic Data Analysis

FIG. 11.2-43. Crossplots of the attributes within (a) the postreservoir zones, (b) the reservoir zones, and (c) the prereservoir
zones labeled in the sections shown in Figure 11.2-42.

Reservoir Geophysics
This in turn requires, strictly, an estimated earth model
in depth that can be used to trace rays associated with
the CRP geometry. Figure 11.2-35 shows a velocitydepth model associated with the CRP stack shown in
Figure 11.2-32. Often, the following procedure is adequate for a velocity-depth model estimation for the
purpose of AVO analysis (Section 9.1):
(a) Interpret a set of time horizons from the CRP stack
(Figure 11.2-32).
(b) Intersect the migration velocity eld from step (g)
of the prestack time migration sequence described
above, which is equivalent to the rms velocity eld
associated with the vertical raypaths, with the time
horizons from step (a) and create a set of horizonconsistent rms velocity proles.
(c) Perform Dix conversion of the rms velocity proles
to derive a set of interval velocity proles.
(d) Convert the time horizons from step (a) to depth
horizons by using the interval velocity proles from
step (c) along image rays.
(e) Combine the depth horizons from step (d) with the
interval velocity proles from step (c) and build a
velocity-depth model as shown in Figure 11.2-35.
Before we proceed with prestack amplitude inversion to derive the AVO attributes, we rst need to verify the existence of amplitude variation with oset on
recorded data by modeling the amplitudes on a CRP
gather coincident with a well location. Actually, we
model a moveout-corrected CMP gather at the well location using a locally at earth model, which is consistent with the earth model associated with the CRP
gather. Figure 11.2-36 shows a sonic and a density log
at a well location that coincides with the CMP location
1134 (Figure 11.2-32). By using the log curves and the
S-wave velocity calculated from the mudrock relation
(equation 11-42), compute the synthetic CMP gather
associated with a horizontally layered earth model at
the well location. Figure 11.2-37 shows a portion of the
CRP gather from prestack time migration as in Figure 11.2-28 and the synthetic CMP gather computed
by using the Zoeppritz equation for P -to-P reection
amplitudes associated with a horizontally layered earth
model (Section L.5). Shown also in this gure is the angle of incidence within the same time window as for the
CRP gather derived from the velocity-depth model in
Figure 11.2-35. It is evident that given the oset range
of 150-3050 m, the maximum angle of incidence at the
target zone (3.1-3.3 s) is nearly 30 degrees. This means
that the AVO attributes based on the Shuey approximation (equation 11-24) can be considered accurate.
A close-up view of the real CRP gather and the synthetic gather of Figure 11.2-37 at the reservoir zone is

1859

shown in Figure 11.2-38. Note that the amplitude variations in the two gathers are fairly consistent. In fact,
examine in Figure 11.2-39 the amplitude variation with
oset at three time levels that coincide with top-gas,
top-oil, and top-water boundaries. We may draw two
conclusions from the comparison of the real and modeled amplitudes as a function of oset shown in Figure
11.2-39. First, note that the Zoeppritz-modeled (red)
and the best-t actual (dotted green) amplitude variation with oset are in very good agreement it appears
that if you scale one AVO curve by a constant, you will
get the other AVO curve. Second, the amplitudes do
indeed vary with oset there is a decrease in amplitude with increasing oset at the top-gas and top-water
boundaries, and there is an increase in amplitudes with
increasing oset at the top-oil boundary.
The top-gas AVO curve shown in Figure 11.2-39 extracted from the real CRP gather is consistent with the
observations made by Rutherford and Williams (1989)
who classied the AVO anomalies associated with a
shale formation over a gas-bearing sand formation. A
Class 1 AVO anomaly indicates a large positive P -toP normal-incidence amplitude (large positive AVO intercept attribute value) and a marked decrease in amplitudes with increasing oset, with a possible phase
change at large osets. According to the RutherfordWilliams classication, the top-gas AVO curve in Figure 11.2-39 can be labeled as Class 1. A Class 2 AVO
anomaly indicates a small positive P -to-P normalincidence amplitude (small positive AVO intercept attribute value) and a marked decrease in amplitudes with
increasing oset, with a possible phase change at smallto-moderate osets. Finally, a Class 3 AVO anomaly
indicates a large negative P -to-P normal-incidence amplitude (large negative AVO intercept attribute value)
and a marked decrease in amplitudes with increasing
oset, typically manifested as bright-spot anomalies.
Another AVO indicator is the dierence between a
near-angle stack (say, up to 15 degrees of angle of incidence as shown in Figure 11.2-37c) and a wide-angle
stack (say, between 15-30 degrees of angles of incidence
as shown in Figure 11.2-37c). If there were no amplitude
variation with oset, there would not be a dierence between the two sections shown in Figure 11.2-40, except
for noise and residual moveout errors.
To calibrate the sections in Figure 11.2-40 to well
data, compute the reectivity series using the two log
curves in Figure 11.2-36. Then, by using a zero-phase
band-limited wavelet with a passband that corresponds
to the signal band associated with the prestack timemigrated data (Figure 11.2-32), compute the zero-oset
synthetic seismogram that contains primaries only. Insert the synthetic seismogram into the CRP stack at the
well location as shown in Figure 11.2-40. Apply constant

1860

Seismic Data Analysis

FIG. 11.2-44. (a) Opacity removal applied to the AVO gradient volume associated with the image volume as in Figure
7.5-31, (b) a magnied view of a portion of (a).

Reservoir Geophysics

1861

FIG. 11.2-45. (a) Opacity removal applied to the AVO intercept volume associated with the image volume as in Figure
7.5-31, (b) a magnied view of a portion of (a).

1862

Seismic Data Analysis

FIG. 11.3-1. A portion of a CMP stack containing a bright spot.

time shift and constant phase shift until a good match


at the reservoir level is attained between the real and
the synthetic data. In the present case, a constant time
shift only, without a constant phase shift, was adequate
to achieve a good match at the reservoir level.
Interpretation of AVO Attributes
We are now ready to perform the actual prestack amplitude inversion to derive the AVO attributes. Figure
11.2-41 shows close-up portions of the AVO attribute
sections at the vicinity of the well location. In each of
the AVO attribute sections, the red represents the high
and the green represents the low attributes values. The
reservoir zone is between the two red horizons in the
CRP stack section shown in Figure 11.2-40. For reference, use Figure 11.2-22 to identify the equations associated with the attributes shown in Figure 11.2-41.
Compare the intercept section (Figure 11.2-41) and
the CRP stack section (Figure 11.2-40), and note that
the former better matches the synthetic attribute derived from the well data. In theory, the intercept attribute is associated with zero-oset, normal-incidence
reectivity. As such, amplitude inversion for acoustic
impedance estimation, whenever possible, should be

made from the intercept attribute derived from the amplitude inversion of CRP gathers.
Now compare each attribute section in Figure 11.241 with the synthetic equivalent at the well location
and note that the match between the two within the
reservoir zone generally is very good. While the intercept section shows the sandstone (green) with shale interbeddings (pale orange) within the reservoir unit, the
gradient section indicates variations in uid saturation
(indicated by the dierent tones of orange) within the
reservoir rocks. The uid saturation can also be inferred
from the Poisson reectivity (equation 11-27) or the
pseudo-Poisson (equation 11-40) section shown in Figure 11.2-41. The orange color in the uid-factor section
represents the gas-saturated reservoir sandstone.
AVO attributes are usually interpreted by crossplotting one attribute against another. Figure 11.2-42
shows three dierent zones postreservoir (blue rectangle), within reservoir (red rectangle) and prereservoir
(green rectangle), labeled on the intercept and gradient
sections. The crossplots of the intercept and gradient
attribute values within each of these zones are shown
in Figure 11.2-43. The slopes of the best-t lines for
the postreservoir (-4.5) and the prereservoir (-5.5) zones
are comparable. Whereas, the slope from the crossplot
is signicantly larger in the reservoir zone (-7.5). The

Reservoir Geophysics
increase in slope within the reservoir zone indicates
an increase in the gradient attribute value, which in
turn, suggests an increase in the change in Poissons
ratio (equation 11-25), hence an indication of the gassaturated reservoir sands.

3-D AVO Analysis


The ultimate goal in seismic interpretation is to derive a
reservoir model. The geometry of the reservoir model is
provided by structural interpretation (Section 7.5), the
result of which is downscaled in the vertical direction
and upscaled in the lateral direction to build the structural framework of the reservoir model. This means that
the reservoir layer is sliced into layers as thin as 1 m in
the vertical direction, and each thin layer is compartmentalized into cells with lateral dimensions as big as
a few hundred meters.
The internal constitution of the reservoir layer itself is specied by using the result of a stratigraphic interpretation (Section 7.5). The interior of the reservoir
layer eventually needs to be populated by the petrophysical properties, such as porosity, permeability, and
uid saturation. When derived from 3-D seismic data,
AVO attributes can be used to infer these properties
within the reservoir zone in inline, crossline, and vertical directions.
The prestack signal processing sequence tailored
for AVO analysis described in this section is applicable directly to 3-D seismic data. To generate the CRP
gathers used in prestack amplitude inversion to derive
the volume AVO attributes, however, you need to follow
the workow for 3-D prestack time migration (Section
7.4). Figures 11.2-44 and 11.2-45 show the gradient and
intercept volumes, respectively, derived from the CRP
gathers as in Figure 7.4-21, following the application
of opacity removal. While the opacity test applied to
the image volume as shown in Figure 7.5-33 enabled
us to identify the presence of a channel system within
the reservoir zone, the opacity tests applied to the attribute volumes shown in Figures 11.2-44 and 11.2-45
indicate characteristics of uid distribution within and
away from the channel. This information is needed to
choose the most desirable drilling locations and thus
better manage the development of the reservoir. Note
from Figure 11.2-45b that the amplitudes labeled in red
that dene the channel also are present along one of the
faults. This suggests that the fault may represent the
migration path for uids. Also note the presence of a
red spot at the center of the channel between the two
faults. This may represent a thicker zone in the ood
plain.

1863

11.3 ACOUSTIC IMPEDANCE ESTIMATION


The goal in amplitude inversion is to obtain a broadband impedance model of a reservoir zone from a
band-limited time-migrated CMP-stacked data. Alternatively, amplitude inversion can be applied to the CRP
stack derived from prestack time migration or the intercept AVO attribute section. The trace amplitudes
on a migrated CMP-stack, CRP stack, or the intercept
section represent the one-dimensional (1-D), primary,
P -wave reectivity series associated with vertical incidence. Care must be given in processing to preserve relative amplitudes and increase vertical and lateral resolution of the seismic data. Reection amplitude variations
on any of these sections may indicate changes in some
rock parameters, such as porosity or uid saturation,
within the reservoir unit.
By considering a sparse-spike reectivity model of
the earth, we rst obtain a broad-band reectivity section. Mathematically, the process involves minimizing
the quantity that is equal to the sum of the absolute
values of the trace amplitudes on the migrated CMP
stack. Computation of the impedance series at a CMP
location then involves simple integration of the broadband reectivity series.
Recall from Section 11.2 that we assume a seismic source that generates a compressional plane wave
propagating down into the earth at an angle from the
vertical. When this incident plane wave encounters a
layer boundary, which is assumed to be at with no
curvature, it is partitioned into four components reected and refracted P - and S-waves. The Zoeppritz
equations (11-12a,b,c,d) describe the amplitudes of the
partitioned wave components. Aside from the layer parameters density, P - and S-wave velocities, the wave
amplitudes depend upon the angle of incidence (Section 11.2). Implicit to CMP stacking, we must assume
that the reected P -wave amplitude variation with angle of incidence measured along the normal-moveout
trajectory associated with the reection event itself on
the CMP gather is negligible. This assumption becomes
critical for shallow targets at large source-receiver osets. A test of the oset range used in stacking for the
time window of interest is crucial for the meaningful interpretation of a poststack amplitude inversion result.
By making the vertical-incidence assumption, we also
assume that there is no P -to-S conversion within the
oset range used in CMP stacking.
The special case of the Zoeppritz equation for vertically incident P -to-P reection amplitude takes a simple form in which the amplitude depends only on the

1864

Seismic Data Analysis

FIG. 11.3-2. A synthetic sonic log section derived from the stacked section in Figure 11.3-1.

impedance values of the layer above and the layer below the reecting boundary (equation 11-9). Based on
this special case of the Zoeppritz equation, computing
the impedance series from a reectivity series involves
a simple integration of the latter. The assumption of
vertical incidence also requires that the stacked section
input to inversion must be migrated.
Results of poststack amplitude inversion must be
viewed within the realm of the underlying critical assumptions we make about seismic amplitudes. Amplitude inversion of poststack time-migrated data should
be limited to earth models with low-relief structures
and stratigraphic targets. Many of the assumptions indicated above are violated in the presence of reectors
with dip and curvature. To account for the dip and curvature eects, at least kinematically, amplitude inversion must be applied to prestack time-migrated data.
Moreover, limitations in the removal of multiple reections, linear and random noise must be taken into consideration when interpreting anomalies observed on an
acoustic impedance section.
Synthetic Sonic Logs
From the discussion on the convolutional model of a
seismic trace in Section 2.1, recall that the starting
point was the sonic log (Figure 2.1-8a). The reection

coecient is dened as the ratio of the reection amplitude to the incident wave amplitude. In terms of acoustic impedance I = , where is rock density and is
the P -wave velocity, the reection coecient c is given
by equation (11-9).
From equation (11-9), the reection coecient is
interpreted as the ratio of the change in acoustic
impedance to twice the average acoustic impedance. If
we assume that density is constant, then equation (119) takes the form
ci =

i+1 i
,
i+1 + i

(11 56)

where i is the sampling index. Therefore, reection coecients can be computed by dierentiating the sonic
log. The inverse process of getting the interval velocities for synthetic sonic logs from the reection coecients, which are assumed proportional to the stacked
trace amplitudes, involves integration. In practice, only
the high-frequency component of the interval velocity
function can be obtained from this inversion. The lowfrequency trend must be obtained from other sources
of information such as conventional velocity analysis or
nearby sonic logs. In many practical situations, there
is insucient well control and the bands of the lowand high-frequency information derived from the seismic data do not overlap. In these cases, problems arise
in merging the various types of information; hence, the
delity of the synthetic sonic logs is degraded.

Reservoir Geophysics
Lindseth (1979) rst introduced the concept of
synthetic sonic logs and used it in the interpretation
of stratigraphic prospects. Figure 11.3-1 illustrates a
stacked section with a bright-spot prospect. The corresponding synthetic sonic log section is shown in Figure 11.3-2 wherein the seismic data are superimposed
as wiggle traces.
Measured sonic logs generally contain much higher
frequencies than seismic data. Integration of seismic
traces to get synthetic sonic logs implies a further lowering of the frequency content. A synthetic sonic log and
a measured log can only be compared when a high-cut
lter is applied to the measured log to make the two
appear to have equivalent bandwidths.
The inversion described here is based on the recursive relation in equation (11-56). Specically, given
values for ci and i , i+1 can be computed. Modelbased, iterative inversion schemes also exist (Cooke and
Schneider, 1983). These begin with an initial impedance
function (specied at a CMP location) from which a
synthetic seismogram is computed. This seismogram is
crosscorrelated with the actual stack trace at that CMP
location. The initial impedance model then is perturbed
until the best match is attained between the estimated
and actual seismic traces. Finally, inversion techniques,
which are based on a certain type of characterization of
the reection coecient series estimated from stacked
data, also exist. For example, the estimated reection
coecient series can be characterized by a sparse spike
series. The sparseness requirement can be satised by
the condition that the sum of the absolute values of the
estimated reection coecients is minimum (Oldenburg
et al., 1983).
Note that regardless of the method, any inversion
process suers from the uncertainty (nonuniqueness)
associated with the frequency components outside the
dominant seismic signal bandwidth. Thus, the results of
the inversion may be questionable for the low-and highfrequency ends of the spectrum at which noise dominates the signal. Constraints (information other than
the seismic data being inverted, such as geologic or well
control) often are incorporated into inversion schemes
used to handle the low-and high-frequency ends of the
spectrum.

Processing Sequence for


Acoustic Impedance Estimation
The parsimonious signal processing sequence described
in Section 11.2 tailored for AVO analysis also should
be used for acoustic impedance estimation. Again, it is
important to preserve the signal bandwidth of therecorded data during processing and attain a broad-

1865

Table 11-4. Data specications for the line used in


acoustic impedance estimation.
Line Length
Shot Spacing
Receiver Spacing
CMP Spacing
Minimum Oset
Maximum Oset
Number of Receivers
Fold of Coverage
Sampling interval

23.2
25
16.67
8.33
80
3064
180
60
2

km
m
m
m
m
m

ms

band spectrum with a at passband for the data input


to poststack amplitude inversion. Also, the processing
sequence must be designed to preserve relative amplitudes in the data. We shall analyze a 2-D marine line
for acoustic impedance estimation. Data specications
for the line are listed in Table 11-4. In addition to the
seismic data, sonic and density logs from a well located
on the line traverse were used in estimating and removing the constant and linear phase associated with the
residual wavelet in the poststack time-migrated CMP
stacked data prior to amplitude inversion.
The processing sequence included the following
steps:
(a) Design a Wiener shaping lter (Section 2.5) to convert the recorded far-eld source signature to its
minimum-phase equivalent. Then, perform signature deconvolution by applying the lter to the
recorded data.
(b) Mute guided waves since they are conned to the
shallow water layer and contain no information
about the subsurface reectivity.
(c) Apply t2 -scaling to compensate for geometric
spreading. As mentioned in Section 11.2, a velocitydependent scaling function should be avoided to
prevent overcorrection of amplitudes of multiples.
(d) Perform predictive deconvolution with unitprediction lag to increase the vertical resolution,
and remove short-period multiples and reverberations. With increasing prediction lag, the amplitude spectrum departs from a at character (Section 2.4). No data-dependent scaling should be applied to ensure preservation of relative amplitudes.
(e) Test CMP stacking for optimum fold (Figure 11.33). The principle criterion is to ensure minimal dependency of reection amplitudes on the angle of
incidence within the time window of interest. By

1866

(f)
(g)

(h)
(i)
(j)

(k)

Seismic Data Analysis

using the NMO-corrected CMP gathers at velocity


analysis locations, derive a spatially varying mute
pattern such that much of the far-oset data are
excluded from stacking. The CMP stack with optimum mute (Figure 11.3-3a) appears to best preserve the atness character of the average amplitude spectrum of the data. The amplitude spectra
and the autocorrelograms imply that there needs
to be further attening of the spectrum within the
passband and removal of reverberations.
Perform velocity analyses at frequent intervals, and
obtain an optimum CMP stack using a time- and
spatially varying mute function.
Apply a frequency-space complex Wiener lter
with unit-prediction lag (f x deconvolution) to
attenuate random noise uncorrelated from trace to
trace (Figure 11.3-4a).
Apply minimum-phase deconvolution to restore
at spectrum within the passband (Figure 11.34b).
Apply time-variant spectral whitening to account
for nonstationarity (Figure 11.3-4c).
Design and apply a time-varying lter, while maintaining the broadest possible signal band. Again,
any data-dependent scaling must not be applied to
preserve relative amplitudes.
Migrate the optimum CMP stacked data using a
phase-shift algorithm. Migration compensates for
the eect of reector curvature on reection amplitudes (equation 11-13a) so that the resulting amplitudes can be related directly to acoustic impedance
variations. The reason for using the phase-shift algorithm is that, compared to nite-dierence or integral methods, it causes the least phase and amplitude errors. Additionally, amplitude inversion often
is applied to data associated with a velocity eld
that varies very mildly in the lateral direction.

Poststack time migration may precede the poststack signal processing, particularly deconvolution. As
a result of migration, we transform a 2-D or a 3-D
waveeld to a 1-D waveeld along the seismic traverse.
Hence, we map the normal-incident energy to verticalincident energy and obtain 1-D zero-oset seismograms
at each CMP location. This satises the underlying assumption for poststack deconvolution.
Figure 11.3-5 shows the test panel for poststack
processing in which migration precedes the signal processing. The sequence includes f x deconvolution,
phase-shift migration (Figure 11.3-5a), spiking deconvolution (Figure 11.3-5b), time-variant spectral whitening (Figure 11.3-5c), and time-variant ltering (Figure
11.3-5d). Finally, a mild (3-trace) Karhunen-Loeve lter was applied to the data to attenuate coherent linear noise and any remaining random noise uncorrelated

from trace to trace (Figure 11.3-5e). The unmigrated


CMP stack with the poststack processing as in Figure
11.3-4 is shown in Figure 11.3-6 and the migrated CMP
stack with the poststack processing as in Figure 11.3-5
is shown in Figure 11.3-7.
Below is a summary of key issues regarding the
processing sequence appropriate for amplitude inversion
applied to poststack time-migrated CMP data.
(a) Avoid f k ltering. This type of ltering invariably causes amplitude distortions. For multiple attenuation, use predictive deconvolution with a long
operator length; the process will be eective at near
osets where periodicity of multiples are nearly
preserved.
(b) Do not apply a velocity-dependent scaling function
to compensate for geometric spreading. The amplitudes of multiples will be overcorrected with a
velocity-dependent scaling function. A t2 -scaling is
favored for data to be used in amplitude inversion.
(c) Use the phase-shift method for migration. Differencing approximations to dierential operators
used in nite-dierence migration induce phase errors and cause amplitude distortions. The usual implementation of the Kirchho migration (equation
H-17, Section H.1) does not include all the terms of
the integral solution (equation H-16) to the scalar
wave equation. As such, the missing terms can inuence the amplitude and phase accuracy of the
resulting migrated data.
(d) You may be required to apply a harsh mute in
stacking. Design a mute function at each velocity analysis location and apply a spatially varying
mute pattern to the data for stacking. The design
criterion is such that reection amplitudes should
be uniform within the oset range included in the
stacking for an event.
(e) You may want to migrate before, and not after,
poststack deconvolution. This is to conform with
the fundamental assumption that deconvolution
acts upon a 1-D zero-oset reection seismogram
recorded over a horizontally layered earth model.
Amplitudes associated with any phantom diractions should not be present on the traces input to
deconvolution.

Derivation of Acoustic Impedance Attribute


As input to poststack amplitude inversion, you may use
one of the following data types:
(text continues on p. 1872)

Reservoir Geophysics

1867

FIG. 11.3-3. A portion of a CMP stack for a test for optimum stacking fold using dierent cable lengths: (a) using an
optimum mute pattern that minimzes angle dependency of seismic amplitudes, (b) using a mute pattern that provides a
full-oset stack, (c) 75 percent near-oset stack, (d) 50 percent near-oset stack, (e) 25 percent near-oset stack, and (f) 17
percent near-oset stack. On top of each panel is the amplitude spectrum averaged over the shot record, and at the bottom
is the autocorrelogram.

1868

Seismic Data Analysis

FIG. 11.3-4. A portion of the CMP stack as in Figure 11.3-3a after: (a) poststack f x deconvolution to attenuate random
noise, (b) spiking deconvolution, and (c) time-variant spectral whitening. On top of each panel is the amplitude spectrum
averaged over the shot record, and at the bottom is the autocorrelogram.

Reservoir Geophysics

1869

FIG. 11.3-5. A portion of the time-migrated CMP stack after: (a) f x deconvolution, (b) spiking deconvolution, (c) timevariant spectral whitening, (d) time-variant ltering, and (e) Karhunen-Loeve ltering. On top of each panel is the amplitude
spectrum averaged over the shot record, and at the bottom is the autocorrelogram.

1870

Seismic Data Analysis

Reservoir Geophysics

1871

1872

Seismic Data Analysis

(a) A time-migrated CMP-stacked section or volume


of data,
(b) A CRP-stacked section or volume of data derived
from prestack time migration,
(c) The intercept AVO attribute section or volume of
data (equation 11-24), or
(d) The P -wave reectivity section or volume of data
(equation 11-37) derived from prestack amplitude
inversion.
Poststack deconvolution and time-variant spectral whitening (Figures 11.3-5b,c) atten the spectrum
within the passband under the minimum-phase assumption. Nevertheless, there may be residual phase remaining in the stacked data; this residual phase typically is
considered to have a constant and a linear phase component. Use of well data makes it possible to estimate
the residual phase. After the completion of the conventional processing phase, the residual wavelet in the
data input to amplitude inversion needs to be estimated
and removed. This requires computing synthetic seismograms at well locations and comparing them with the
stacked traces to determine constant-phase and linearphase components associated with the residual phase
left in the migrated data input to amplitude inversion.
(a) To start with, inspect and edit available sonic and
density logs, and use check-shot information for
depth-to-time conversion of the sonic logs (Figure
11.3-8a).
(b) Compute the acoustic impedance curve and obtain
synthetic seismograms (Figure 11.3-9a) using zerophase wavelets with a range of bandwidths that
cover the passband of the time-migrated stacked
data.
(c) Apply a range of constant phase rotations (Figure 11.3-9b) to the traces from the time-migrated
CMP stacked section at the vicinity of the well
and compare the results with the synthetic seismograms. In the present case study, we observed
that the time-migrated stack did not require any
phase correction within the upper portion of the
zone of interest (2.5-3.5 s). Nevertheless, a better
match was observed between the well data and the
surface seismic data with a constant phase rotation
of 90 degrees within the lower portion of the zone
of interest (3.5-4.5 s). Figures 11.3-10 and 11.3-11
show portions of the time-migrated stacked section
as in Figure 11.3-7 with the application of a constant phase rotation of 90 degrees.
(d) Compute a sparse-spike reectivity model of the
earth by a constrained L1 -norm minimization applied to the time-migrated sections in Figures 11.310b and 11.3-11b.

(e) Apply a mild (3-trace) Karhunen-Loeve lter to


the reectivity sections to remove any spurious, geologically implausable variations in the reectivity
model (Figures 11.3-12a and 11.3-13a).
(f) Construct the acoustic impedance model from the
broad-band sparse-spike reectivity model by integrating the reectivity series at each CMP location
(Figures 11.3-12b and 11.3-13b) within the zone of
interest (2.5 - 4.5 s).
(g) Repeat steps (d), (e), and (f) for the phase-rotated
sections in Figures 11.3-10c and 11.3-11c. Results
are shown in Figures 11.3-14 and 11.3-15.
We applied a range of constant phase rotations
to a group of traces at two dierent locations on
the migrated CMP stack and computed the acoustic
impedance panels shown in Figures 11.3-16 and 11.317. Comparing these panels with the actual impedance
curve at the well location (Figure 11.3-8b), we observed
that the match between the two impedances in the upper portion of the zone of interest (2.5-3.5 s) is satisfactory without phase rotation applied to the data, while
a better match is attained with a 90-degree phase rotation in the lower portion of the zone of interest (3.5-4.5
s). Any discrepancies may be attributed primarily to
inaccuracies in the check-shot survey information.
It is important to evaluate the results of amplitude inversion with a cautious perspective. Specically,
one cannot and should not infer individual reservoir
properties pore pressure, conning pressure, porosity, permeability, and uid saturation, from amplitude
inversion. One can only hope to infer a composite effect associated with these properties. Given additional
borehole information, however, one may be able to infer
that a high impedance value may correspond to, for example, low porosity. In brief, results of amplitude inversion should be used as auxiliary information in support
of other tools used in exploration and development of
hydrocarbons.
3-D Acoustic Impedance Estimation
As for the AVO attributes, when derived from 3-D seismic data, an acoustic impedance attribute volume can
be used to infer lithology or porosity within a reservoir
zone. We shall analyze a land 3-D data set for acoustic impedance estimation. Data specications for the
3-D seismic data are listed in Table 11-5. The fold of
coverage over the survey area is fairly uniform (Figure
11.3-18). In addition to the seismic data, sonic and density logs from a well from within the survey area were
used in estimating and removing the constant and linear phase associated with the residual wavelet in the
(text continues on p. 1885)

Reservoir Geophysics

1873

FIG. 11.3-8. (a) A portion of the sonic log that spans the time window of interest at one well location on the line shown in
Figure 11.3-7 after check-shot correction and depth-to-time conversion; (b) acoustic impedance curves derived from the log in
(a), assuming that density is constant, with a range of constant phase rotations.

1874

Seismic Data Analysis

FIG. 11.3-9. (a) Synthetic seismograms derived from the impedance curve in Figure 11.3-8 and using a range of pass-bands,
with normal (n) and reverse (r) polarity displays; (b) traces from the migrated CMP stacked section in Figure 11.3-7 in the
neighborhood of the well location with the sonic log shown in Figure 11.3-8a with a range of constant phase rotations.

Reservoir Geophysics

1875

FIG. 11.3-10. First portion of: (a) unmigrated CMP stack as in Figure 11.3-6; (b) migrated CMP stack as in Figure 11.3-7;
(c) migrated CMP stack with a constant phase rotation of 90 degrees. A display gain has been applied to all three sections.

1876

Seismic Data Analysis

FIG. 11.3-11. Second portion of: (a) unmigrated CMP stack as in Figure 11.3-6; (b) migrated CMP stack as in Figure 11.3-7;
(c) migrated CMP stack with a constant phase rotation of 90 degrees. A display gain has been applied to all three sections.

Reservoir Geophysics

1877

FIG. 11.3-12. (a) Broad-band sparse-spike reectivity section derived from the wide-band time-migrated section shown in
Figure 11.3-10b; (b) acoustic impedance section derived from (a).

1878

Seismic Data Analysis

FIG. 11.3-13. (a) Broad-band sparse-spike reectivity section derived from the wide-band time-migrated section shown in
Figure 11.3-11b; (b) acoustic impedance section derived from (a).

Reservoir Geophysics

1879

FIG. 11.3-14. (a) Broad-band sparse-spike reectivity section derived from the wide-band time-migrated section shown in
Figure 11.3-10c; (b) acoustic impedance section derived from (a).

1880

Seismic Data Analysis

FIG. 11.3-15. (a) Broad-band sparse-spike reectivity section derived from the wide-band time-migrated section shown in
Figure 11.3-11c; (b) acoustic impedance section derived from (a).

Reservoir Geophysics

1881

FIG. 11.3-16. Part 1: Portion of the acoustic impedance section in Figure 11.3-12b around CMP 370 with constant phase
rotations applied.

1882

Seismic Data Analysis

FIG. 11.3-16. Part 2: Portion of the acoustic impedance section in Figure 11.3-13b around CMP 370 with constant phase
rotations applied.

Reservoir Geophysics

1883

FIG. 11.3-17. Part 1: Portion of the acoustic impedance section in Figure 11.3-12b around CMP 450 with constant phase
rotations applied.

1884

Seismic Data Analysis

FIG. 11.3-17. Part 2: Portion of the acoustic impedance section in Figure 11.3-13b around CMP 450 with constant phase
rotations applied.

Reservoir Geophysics

1885

FIG. 11.3-18. Fold of coverage map of the land 3-D survey data used in the acoustic impedance case study presented in
Section 11.3.

time-migrated CMP stacked data prior to amplitude


inversion.
Note the high level of ground-roll energy on the
selected shot records with the display gain shown in
Figure 11.3-19. The noise characteristics vary from shot
to shot. The processing sequence included the following
steps:

(a) Apply t2 -scaling to compensate for geometric


spreading. Figure 11.3-20a shows a raw shot record
and Figure 11.3-20b shows the same record after
the application of t2 -scaling.
(b) Perform spiking deconvolution (Figure 11.3-21a).
(c) Note from the spectral estimate shown in Figure
11.3-22 that deconvolution alone does not atten
(text continues on p. 1896)

1886

Seismic Data Analysis

FIG. 11.3-19. Part 1: Selected shot records from the 3-D survey data associated with the case study presented in Section
11.3. (Data courtesy Talisman Energy.)

Reservoir Geophysics

1887

FIG. 11.3-19. Part 2: Selected shot records from the 3-D survey data associated with the case study presented in Section
11.3. (Data courtesy Talisman Energy.)

1888

Seismic Data Analysis

FIG. 11.3-20. (a) A raw shot record from the 3-D seismic data associated with the case study presented in Section 11.3, (b)
after geometric spreading correction. (Data courtesy Talisman Energy.)

Reservoir Geophysics

1889

FIG. 11.3-21. (a) The same shot record as in Figure 11.3-20b after deconvolution, and (b) time-variant spectral whitening.
(Data courtesy Talisman Energy.)

1890

Seismic Data Analysis

FIG. 11.3-22. Amplitude spectrum of the shot record in (a) Figure 11.3-20a, (b) Figure 11.3-20b, (c) Figure 11.3-21a, and
(d) Figure 11.3-21b.

Reservoir Geophysics

1891

FIG. 11.3-23. (a) An inline section from the image volume derived from 3-D poststack time migration, (b) amplitude
spectrum of (a).

1892

Seismic Data Analysis

FIG. 11.3-24. Part 1: Selected inline sections from the image volume derived from 3-D poststack time migration.

Reservoir Geophysics

1893

FIG. 11.3-24. Part 2: Selected inline sections from the image volume derived from 3-D poststack time migration.

1894

Seismic Data Analysis

FIG. 11.3-24. Part 3: Selected inline sections from the image volume derived from 3-D poststack time migration.

Reservoir Geophysics

1895

FIG. 11.3-24. Part 4: Selected inline sections from the image volume derived from 3-D poststack time migration.

1896

Seismic Data Analysis

Table 11-5. Data specications for the 3-D seismic


data used in acoustic impedance estimation.
Survey size
Inline dimension
Crossline dimension
Receiver line spacing
Bin size
Number of inlines
Number of crosslines
Number of bins
Fold of coverage
Sampling interval

(d)
(e)
(f)
(g)
(h)
(i)
(j)
(k)

27.3
7.35
3.75
60
30 30
124
245
30,380
16
2

km2
km
km
m
m

ms

the spectrum. Apply time-variant spectral whitening (Figure 11.3-21b) to attain a at spectrum
within the signal passband (Figure 11.3-22d) and
attenuate the ground-roll energy.
Sort the data into common-cell gathers and perform velocity analysis at coarse grid spacing.
Create a 3-D stacking velocity eld and apply NMO
correction to the cell gathers.
Apply residual statics corrections (Section 7.2) and
inverse NMO correction using the velocity eld
from step (e).
Perform velocity analysis at tight grid spacing and
create a new 3-D velocity eld.
Apply NMO correction to the cell gathers from step
(f), mute and stack.
Perform poststack deconvolution and a wide bandpass lter.
Apply f x deconvolution (Section 6.5) to attenuate random noise uncorrelated from trace to trace.
Finally, perform 3-D poststack phase-shift migration. Note from Figure 11.3-23 that the processing
sequence described above yields the spectral shape
of the data that we wish to input to poststack amplitude inversion a broadband spectrum with
nearly a at passband.

Figure 11.3-24 shows selected inlines from the image volume derived from 3-D poststack time migration.
Following the principle of parsimony in processing for
AVO analysis and acoustic impedance estimation, no
DMO correction was deemed necessary in the present
case study since the subsurface geology can be described
by a horizontally layered earth model with no signicant
faulting or structural distortions.
The next phase in the analysis involves the estimation and removal of the residual phase contained in the
time-migrated volume of data. Figure 11.3-25 shows a
sonic and a density log measured at a well located at
the intersection of inline 76 and crossline 79. Compute

the acoustic impedance and the reectivity series; then,


use a zero-phase wavelet with a passband comparable to
the signal bandwidth of the seismic data to compute the
zero-oset, vertical-incidence synthetic seismogram also
shown in Figure 11.3-25. Note from Figure 11.3-26 that
the sonic log inserted into a portion of inline 76 at the
well location exhibits a very good match with the timemigrated data within the reservoir level indicated by
the arrow. To achieve this match, a constant time shift,
which is equivalent to a linear phase shift, was applied
to the sonic log. A similarly good match is observed in
Figure 11.3-27 between the synthetic seismogram and
the time-migrated data. To achieve this match, in addition to the constant time shift, a 90-degree phase
rotation was applied to the synthetic seismogram. To
remove the residual phase in the migrated data set and
match it to the zero-phase synthetic seismogram, you
therefore need to apply a 90-degree phase rotation to
the former.
Figure 11.3-28a shows inline 76 of the image volume from 3-D poststack time migration, and Figure
11.3-28b shows the same section after the 90-degree
phase rotation. The phase-rotated image volume is then
used in amplitude inversion to create a sparse-spike
reectivity volume. Figure 11.3-28c shows inline 76
from the sparse-spike reectivity volume. Finally, each
sparse-spike reectivity trace is integrated to estimate
the acoustic impedance attribute (Figure 11.3-28d). Selected inlines from the acoustic impedance volume are
shown in Figure 11.3-29. These inline traverses coincide
with those shown in Figure 11.3-24 from the image volume derived from 3-D poststack time migration before
the application of phase rotation.
The acoustic impedance volume is then interpreted
to identify the spatial distribution of high-impedance
and low-impedance areas at the reservoir level. First,
pick the time horizon at the reservoir level from the image volume and isolate a horizon-consistent thin slab of
subvolume. Then, apply transparency to the subvolume
to obtain the interpretation result shown in Figure 11.330. The red represents the areas of high-amplitude reections at the reservoir level. Apply the same interpretation strategy to the acoustic impedance volume and
obtain a consistent result as shown in Figure 11.3-31.
The orange tones represent the zones of high acoustic
impedance contrast.

Instantaneous Attributes
Aside from the AVO attributes, the instantaneous attributes can be helpful in inferring reservoir parameters.
When considered as an analytic signal (in the mathematical sense), a seismic trace can be expressed as a
(text continues on p. 1906)

Reservoir Geophysics

1897

FIG. 11.3-25. Log data measured at well location Inline 76 - Crossline 79: (a) the sonic log, (b) the density log, (c) acoustic
impedance, (d) the reectivity series computed from (c) and used in creating (e), and (e) synthetic seismogram using a
band-limited wavelet.

1898

Seismic Data Analysis

FIG. 11.3-26. A portion of Inline 76 from the image volume derived from 3-D poststack time migration with the sonic log
as in Figure 11.3-25a inserted at well location, Crossline 79.

Reservoir Geophysics

1899

FIG. 11.3-27. A portion of Inline 76 as in Figure 11.3-26 from the image volume derived from 3-D poststack time migration
with the synthetic seismogram as in Figure 11.3-25e inserted at well location, Crossline 79.

1900

Seismic Data Analysis

FIG. 11.3-28. Cross-sections along inline 76 from (a) the image volume derived from 3-D poststack time migration, (b) the
image volume after 90-degree phase rotation, (c) the sparse-spike reectivity volume, and (d) the acoustic impedance volume.

Reservoir Geophysics

FIG. 11.3-29. Part 1: Selected inline sections from the acoustic impedance volume.

1901

1902

Seismic Data Analysis

FIG. 11.3-29. Part 2: Selected inline sections from the acoustic impedance volume.

Reservoir Geophysics

1903

FIG. 11.3-30. A 3-D view of the detrital sand level at approximately 850 ms from within the image volume derived from
3-D poststack time migration as in Figure 11.3-24. Red represents high amplitudes.

1904

Seismic Data Analysis

FIG. 11.3-31. A 3-D view of the detrital sand level at approximately 850 ms from within the acoustic impedance volume
derived from sparse-spike inversion as in Figure 11.3-29. Red represents high impedance values.

Reservoir Geophysics

1905

FIG. 11.3-32. Instantaneous attributes derived from the stacked section in Figure 11.3-1: (a) reectivity strength, (b) phase,
(c) frequency, and (d) smoothed frequency.

1906

Seismic Data Analysis

FIG. 11.4-1. Vertical seismic proling geometry. (a) Raypaths and (b) associated traveltimes (see text for details). Static
correction amounts to mapping traveltime associated with raypath ABC to traveltime associated with raypath ABC + CD;
the NMO correction amounts to mapping traveltime associated with raypath ABCD to traveltime associated with raypath
2DE.

complex function (Taner, 1978). An analytic signal is


expressed by a time-dependent complex variable u(t)
u(t) = x(t) + iy(t),

(11 57)

where x(t) is the recorded trace itself and y(t) is its


quadrature. The quadrature is a 90-degree phase-shifted
version of the recorded trace. It is obtained by taking
the Hilbert transform of x(t) (Bracewell, 1965)
1
y(t) =
x(t).
t

(11 58)

When substituted into equation (11-57), we have


1
x(t).
u(t) = x(t) (t) + i
t

(11 59)

Thus, to get the analytic signal u(t) for a seismic trace


x(t), the complex operator [(t) + i/t] is applied to the
trace. When analyzed in the Fourier transform domain,
this operator is zero for negative frequencies. There-

fore, the complex trace u(t) does not contain negative


frequency components.
Once u(t) is computed, it can be expressed in polar
form as
u(t) = R(t) exp i(t) ,

(11 60)

where
R(t) =

x2 (t) + y 2 (t)

(11 61a)

and
(t) = arctan

y(t)
.
x(t)

(11 61b)

R(t) represents instantaneous amplitude and (t) represents instantaneous phase at time t. In practice, instantaneous phase is computed by rst taking the logarithm
of both sides of equation (11-60)
ln u(t) = ln R(t) + i(t)

(11 62)

Reservoir Geophysics
and extracting the imaginary part
(t) = Im ln u(t) ,

(11 63)

where Im denotes imaginary.


The instantaneous frequency (t) is the temporal
rate of change of the instantaneous phase function
d(t)
.
(11 64a)
dt
When the derivative of equation (11-63) is taken with
respect to time,
(t) =

d(t)
1 du(t)
.
= Im
dt
u(t) dt

(11 64b)

For practical implementation, equation (11-64b) is written as the dierence equation


2
ut utt
t =
Im
.
(11 65)
t
ut + utt
The instantaneous amplitude measures the reectivity strength, which is proportional to the square root
of the total energy of the seismic signal at an instant
of time. The instantaneous phase is used to emphasize
the continuity of events on a seismic section. The temporal rate of change of the instantaneous phase is the
instantaneous frequency. The instantaneous frequency
may have a high degree of variation, which may be related to stratigraphy. However, it also may be dicult
to interpret all this variation. Therefore, the instantaneous frequency values often are smoothed in time.
The instantaneous measurements that are related
to an analytic signal are associated with an instant of
time, rather than an average over a time interval. These
measurements are reliable when the seismic signal is
recorded and processed so that the CMP stack closely
represents the subsurface. In other words, to deduce any
stratigraphic meaning from the seismic data before estimating the instantaneous parameters, the amplitude
and frequency content of the seismic signal must be
preserved in each processing step. Any variation in the
shape of the basic waveform that is not attributable to
the subsurface geology must be eliminated. Multiples
and all types of random noise limit the reliability of the
results.
Reectivity strength is an eective tool to identify bright and dim spots. Phase information is useful in delineating such interesting features as pinchouts,
faults, onlaps, and prograding reections. Instantaneous
frequency information can help to identify condensate
reservoirs and gas reservoirs, which tend to attenuate
high frequencies.
Instantaneous attributes often are displayed in
color for interpretation. Figure 11.3-32 shows the instantaneous attributes that correspond to the seismic
section in Figure 11.3-1. The seismic data are displayed

1907

on top of the attributes as wiggle traces. Note the distinctive anomaly on the instantaneous amplitude section and enhancement of the continuity of reections
on the instantaneous phase section.

11.4 VERTICAL SEISMIC PROFILING


Vertical seismic proling data often provide more reliable correlation of well control to seismic data than
synthetic seismograms derived from sonic logs. There
are two reasons for this. First, the VSP data have a signal bandwidth closer to the seismic data than the sonic
logs. More importantly, VSP data usually are not as
sensitive to borehole conditions such as washouts.

VSP Acquisition Geometry


Acquisition of VSP data involves a surface source that
is located either close to the well head (zero-oset case)
or away from the well head (oset VSP) and a geophone
in the well bore. Several traces are recorded at the same
geophone depth, then edited and summed. (Repeatable
sources such as vibrators or air guns are used.) The
geophone then is moved to a new depth location and
recording is repeated. The resulting section is a prole
displayed in depth and time. A comprehensive reference
on VSP is the book by Hardage (1983).
Figure 11.4-1a shows VSP acquisition geometry.
Consider a few of the raypaths that are intercepted by
the receivers down the borehole: the direct arrivals from
shot to receivers AC and AE, reection ABC, and refraction ABF. Each receiver location yields a trace on
the VSP that is in the depth-time domain (Figure 11.41b). Note that trace C has both the direct arrival (1)
and the reection from the rst interface (3). The direct
arrival (1 on trace C) and the refraction (4 on trace F)
have only downgoing paths; therefore, they are called
downgoing waves. On the other hand, reection path
ABC has a nal upcoming segment BC (Figure 11.41a) and therefore constitutes an upcoming wave (Figure
11.4-1b). Note that direct arrival 2 on trace E coincides
with a reection arrival, provided the receiver is situated on the interface that causes the reection. At this
location, the upcoming and downgoing waves coincide
(arrival 2 on trace E in Figure 11.4-1b).

Processing of VSP Data


We now discuss the basic steps in VSP processing. After some trace editing, VSP processing starts with the

1908

Seismic Data Analysis

separation of the downgoing waves from the upcoming


waves (reections). One separation technique is based
on f k ltering. Examine the zero-oset VSP in Figure 11.4-2a and note that the upcoming and downgoing
waves have opposite dips. Because of this, each wave
type should map in a dierent half plane in the f k domain. Hence, downgoing waves can be suppressed with
an f k dip lter, thereby leaving only the reection
and associated multiples that constitute the upcoming
waves (Figure 11.4-2b).
A VSP data set may not always have uniform receiver sampling in depth a prerequisite for f k ltering. The problems of edge eects and amplitude smearing often are observed on data after f k ltering (Section 6.2).
An alternate approach to extracting upcoming
waves is to use median ltering (Hardage, 1983). To
start, apply rst-arrival time shifts to traces in a VSP
data set VSP(z,t) to atten the downgoing waves. (First
arrivals now are at t = 0.) Then apply a median lter to
each horizontal array of samples, V SP (z, t = constant).
Median ltering is best explained by an example. Consider the following array of numbers: (-1,2,1,2.5,1.5).
Reorder its elements from small to large values: (1,1,1.5,2,2.5). The median of the series then is the middle sample, 1.5. Median ltering rejects noise bursts
and any event that is not at. Apply the median lter
to the VSP data set with attened downgoing waves
to yield the downgoing waves. This result can be subtracted from the input to obtain the upcoming waves.
The last step involves unattening the data.
The next VSP processing step involves datuming
all receivers to the well head (D in Figure 11.4-1a).
From Figure 11.4-1, this static correction is the same
as correcting each trace by an amount that is equal to
the one-way traveltime down to the corresponding receiver location. For example, trace C is corrected by an
amount equal to the traveltime associated with raypath
DC.
The static corrections are followed by deconvolution and ltering (Figure 11.4-2c). In principle, the deconvolution operators can be designed from downgoing
or upcoming waves. These deconvolution operators then
are applied to traces of the upcoming wave prole. The
common practice is to use the downgoing waves to design the deconvolution operators. This is because downgoing waves on a VSP record are much stronger than
upcoming waves. Hence, designing deconvolution operators from downgoing waves has an advantage in that
the operators are based on stronger signal, with more
emphatically represented multiples (Hardage, 1983).
The last step involves stacking the traces in Figure
11.4-2c. Stacking normally includes a narrow corridor

along the region in which upcoming and downgoing


waves coincide. The resulting trace, repeated a few
times, is shown in Figure 11.4-2d. To a large degree, corridor stacking prevents the multiples that do not merge
with the downgoing wave path from being stacked in.
The trace in Figure 11.4-2d can be considered an alternative to a zero-oset synthetic seismogram derived
from the sonic log; thus, it can be compared to the CMP
stack (not shown here) at the well location.
Figure 11.4-3a shows raw data from an oset VSP
data set. By using the median ltering scheme described
earlier, downgoing waves are extracted from the raw
data (Figure 11.4-3b). The subtraction of downgoing
waves from original data (followed by another median
ltering for suppression of noise bursts) yields upcoming
waves (Figure 11.4-3c). The upcoming waves then are
deconvolved (Figure 11.4-3d) using operators designed
from the downgoing waves of Figure 11.4-3b. This step
normally is followed by static corrections. For nonzerooset data, we also must correct for moveout resulting
from the oset separation between the well head and the
shot location. From Figure 11.4-1, this dynamic correction involves mapping the traveltime associated with
raypath ABCD to 2DE. An NMO-corrected upcomingwave prole can be compared with the surface seismic
at the well location, provided the subsurface consists of
horizontal layers with no lateral velocity variations.

VSP-CDP Transform
When there are dipping interfaces, the upcoming-wave
prole needs to be migrated; that is, the energy must
be mapped to the actual subsurface reection points.
This is true even for zero-oset VSP data. A ray-tracing
procedure for reector mapping is illustrated in Figure
11.4-4. Note that reection points D, E, and F have different lateral displacements OA, OB, and OC, respectively, from borehole Oz (Figure 11.4-4a). However, upcoming wave energy from all three reection points is
recorded on the same VSP trace at the receiver location
R. The reection times RG, RH, and RK (Figure 11.44b) are associated with raypaths SDR, SER, and SFR,
respectively. Mapping this energy to reection points
involves a coordinate transformation (Wyatt and Wyatt, 1981; Cassell et al., 1984). In this transformation,
the amplitudes on a single VSP trace are mapped onto
several traces on the x t plane, where x is the lateral
distance of reection points from the borehole (Figure

Reservoir Geophysics

1909

FIG. 11.4-2. Zero-oset VSP data at various stages of the processing sequence from left to right: (a) raw data, (b) upcoming
waves, (c) after statics corrections followed by deconvolution and band-pass ltering, (d) corridor stack. Here, TT is a tube
wave traveling through the borehole. (Data courtesy Amoco Europe and West Africa, Inc.)

1910

Seismic Data Analysis

FIG. 11.4-3. (a) Unprocessed oset VSP data, (b) downgoing waves, (c) upcoming waves, (d) upcoming waves after deconvolution, (e) VSP-CDP transform of the upcoming wave prole in (d) inlaid to the migrated surface seismic section for
comparison. (Adapted from Alam and Millahn, 1986; data courtesy Shell, U.K.)

11.4-4b). The event times RG, RH, and RK are mapped


onto two-way vertical times AL, BM, and CN, respectively. These vertical times are associated with raypaths AD, BE, and CF in Figure 11.4-4a. The resulting
(x,t) section consists of traces similar to the traces of a
migrated zero-oset section. Hence, the described raytracing procedure often is called VSP-CDP transformation.
The VSP-CDP transform of the data in Figure
11.4-3d is shown in Figure 11.4-3e (Alam and Millahn,
1986). Comparison with the migrated surface seismic

section at the well location indicates a good correlation


of events. The dierence in frequency content is partly
attributed to dierences in processing these two sections
and is partly attributed to less high-frequency attenuation eects because of the shorter traveltimes associated
with VSP recording.
The VSP-CDP transformation requires knowledge
of the velocity-depth model around the borehole, since
we must determine the location of the reection points
in the subsurface to perform mapping. The velocitydepth model can be derived using an iterative approach

Reservoir Geophysics

1911

FIG. 11.4-4. (a) Source-receiver geometry for oset VSP; (b) VSP-CDP transformation: Traveltimes RG, RH, and RK,
associated with raypaths SDR, SER, and SFR are mapped to traveltimes AL, BM, and CN, associated with the two-way
vertical raypaths 2AD, 2BE, and 2CF . Also note that the amplitudes on the VSP trace are positioned on traces after
transformation with x-coordinate values the same as those of the reection points OA, OB, and OC.

(Cassell et al., 1984). Starting with an initial velocitydepth model and the recording geometry for the VSP
data, traveltimes for upcoming waves are computed.
When these estimated traveltimes are compared with
the observed traveltimes, discrepancies are noted and
the velocity-depth model is modied accordingly. The
process is repeated until a good match is attained between the estimated and observed traveltimes.
Finally, note that the VSP-CDP transform is not
exactly a migration process. It handles neither diractions nor curved interfaces. To handle these features,
VSP data must be migrated (Dillon and Thomson,
1983). The VSP geometry is like the geometry of a
common-shot gather, except the shot axis is perpendicular to the receiver axis. Migration of VSP data can
be viewed as mapping amplitudes along semielliptical
trajectories with their focal points being the source and
receiver locations. Superposition of all these trajectories yields the migrated section. The aperture width for
VSP data often is inadequate to obtain a migrated section without much smearing.

11.5 4-D SEISMIC METHOD


There is a strong resemblence between the techniques
used in clinical medicine and geophysical prospecting.

Here, I will refer to this resemblence within the context


of the 4-D seismic method. A heart patient is monitored
from year to year by the patients cardiologist to observe
and detect changes in the parameters that describe the
heart itself and its condition to sustain the patients life,
and the composition of the patients blood. By using
an echocardiogram derived from ultrasound waves, the
cardiologist measures the size of the heart and observes
whether the valves have any leakage. By recording an
electrocardiogram, the cardiologist observes the systolic
pressure associated with the rythmical contraction of
the heart during which the blood is pumped out and
the diastolic pressure associated with the rythmical dilatation of the heart during which the blood is pumped
in. It is not just the hearts parameters themselves that
are important to the cardiologist, but also the critical
changes in those parameters from year to year during
the patients life span. Eventually, the cardiologist uses
the data associated with the patients medical history
to judge whether or not a surgical intervention is required, at which time an angiogram is taken to make
direct observations of the heart to provide the necessary
instantaneous information to the surgeon. When evaluating the historical data, the cardiologist is cognizant
of the fact that the hearts paramaters are directly inuenced by the evolving technologies used in taking the

1912

Seismic Data Analysis

measurements, and the interpretation of the measurements is inuenced by the changing medical knowledge
and experience of the cardiologist.
Similarly, the reservoir geophysicist uses the 3-D
seismic method combined with the direct observations
made at well locations to monitor the reservoir conditions that are crucial for optimum development of the
eld. The objective in optimum eld development is to
lengthen the life span of the eld, prevent water invasion, and recover as much hydrocarbon as possible. By
recording 3-D seismic data over the eld at various time
intervals, which may be from months to years, we introduce the fourth dimension to the the analysis of the data
calendar time, thus the term 4-D seismic method to
describe the time-lapse 3-D seismic exploration.
What type of reservoir parameters can we monitor
using the 4-D seismic method? Known applications of
the 4-D seismic method are summarized below ( Lumley
et al., 1994; Lumley, 1995a,b; Ecker, 1999; MacLeod et
al., 1999).
(a) Monitoring the spatial extent of the steam front following in-situ combustion or steam injection used
for thermal recovery,
(b) Monitoring the spatial extent of the injected water
front used for secondary recovery,
(c) Imaging bypassed oil,
(d) Determining ow properties of sealing or leaking
faults, and
(e) Detecting changes in oil-water contact.
A demonstrative example of the 4-D seismic
method is shown in Figure 11.5-1 (MacLeod et al.,
1999). The two seismic sections along the same inline
traverse have been extracted from the image volumes
derived from 3-D poststack time migrations of two timelapse 3-D seismic data. One survey was conducted in
1989 before the production commenced in the eld and
the other survey was conducted in 1998 sometime after the production was started. The oil-water contact
(OWC) is distinctly visible in the 1989 section, while
it is not apparent in the 1998-section. Additionally, the
top-sand reector is stronger in the 1998 section. This
may be attributed to the increased impedance contrast
between the overlying shales and the reservoir sands
which have higher water saturation as a result of continuing production.
Figure 11.5-2 shows another application of the 4-D
seismic method to steam ooding of a reservoir for thermal recovery. In this eld study, six time-lapse 3-D seismic surveys were conducted (Lumley, 1995b). The rst
survey at time T 1 was conducted prior to steam injection, while the subsequent surveys were conducted after

the steam injection was started. The red, near-circular


feature on the time slices correspond to the spatial extent of the injected steam; note how the steam front
expands further in the northwesterly direction with increasing calendar time.

Processing of 4-D Seismic Data


Just as in the case of the medical example given above,
processing, inversion, and interpretation of 4-D seismic
data are inuenced by the evolving technologies in 3-D
seismic exploration. The dierent vintages of 3-D seismic data that are used in a 4-D seismic project are
often recorded with dierent vessels, source and cable
geometries, and source and receiver types and arrays.
In fact, some 4-D seismic projects may involve, say, two
time-lapse 3-D surveys one conducted using the conventional streamer cable and the other conducted using
the ocean-bottom 4-C technique (next section). The 3-D
surveys most likely would be conducted using dierent
recording directions and bin sizes. Figure 11.5-3 shows
the base maps for two time-lapse 3-D surveys with different recording directions and bin sizes (Rickett and
Lumley, 1998). The data associated with the 1979 survey and the 1991 survey were recorded with a 34-degree
dierence in grid orientation. Also, the bin size for the
1979-survey was 80 27.5 m, whereas the bin size for
the 1991 survey was 12.5 12.5 m. Additionally, these
3-D seismic data sets most likely would be processed
dierently not only the processing sequences would
be dierent but also the processing parameters. Figure
11.5-4 (Rickett and Lumley, 1998) shows a section and
a time slice from each of the two time-lapse 3-D surveys
which are referred to in Figure 11.5-3. The 1991 survey
data have produced a more accurate image of the salt
ank.
Hence, the time-lapse 3-D data sets used in a 4-D
seismic project need to be cross-equalized prior to the
interpretation of the results. Cross-equalization involves
the following steps (Rickett and Lumley, 1998).
(a) Align the grids of the time-lapse data to a common grid orientation. In many cases of 4-D seismic
projects, grid alignment and subsequent steps in
cross-equalization are applied to poststack data. As
such, grid alignment may be achieved by remigrating the poststack data to a specied common grid
orientation. If you have access to prestack data,
one way to align the grids of the time-lapse data is
by crossline migration (Section 7.4), the output of
which would be common-azimuth gathers.

Reservoir Geophysics

1913

FIG. 11.5-1. Time-lapse seismic data: (a) preproduction (1989), and (b) postproduction (1998). (MacLeod et al., 1999; gure
courtesy MacLeod et al., Chevron and Schlumerger Geco-Prakla; data courtesy Chevron.)

(b) Apply spectral balancing to the time-lapse 3-D data


to account for the dierences in the spectral bandwidth and shape. Figure 11.5-5 shows the amplitude spectra computed from the 1979 and 1991
survey data shown in Figure 11.5-4. Note the signficant dierences in the shape and bandwidth of the
spectra before cross-equalization. These dierences
have been minimized by cross-equalization.
(c) Derive amplitude gain curves from the time-lapse
3-D data based on trace envelopes, and apply the
gain curves for amplitude balancing.
(d) Estimate static shifts between the time-lapse data
traces and apply them to eliminate vertical time
diferences.
(e) Examine and determine dierences in event positioning in the migrated data volumes associated
with the time-lapse 3-D surveys. Eliminate the differences in event positioning by residual migration.

Seismic Reservoir Monitoring


Figure 11.5-6 shows a dierence section from the
time-lapse data as in Figure 11.5-4 following crossequalization. This dierence section exhibits a strong
amplitude anomaly at the reservoir level situated at
the salt ank. Such an amplitude dierence may be attributed to changes in the reservoir conditions as a result of production (Rickett and Lumley, 1998). Because
of a wide range of factors associated with acquisition
and analysis of the 4-D data, in addtion to the dierence data volume, the individual data volumes themselves are also visualized and interpereted.
The example of cross-equalization shown in Figure 11.5-7 relate to a steam injection project. Note the
dierences in the time slices from the image volumes
associated with the 1996 survey and 1997 survey before

1914

Seismic Data Analysis

FIG. 11.5-2. Time lapse-seismic data: six generations of data displayed side by side. (Lumley, 1995b; gure courtesy Lumley,
4th Wave Imaging and Chevron; data courtesy Chevron.)

Reservoir Geophysics

1915

FIG. 11.5-3. Survey geometries of two time-lapse 3-D seismic data as in Figure 11.5-4: (a) the grid orientations, and (b) grid
sizes. See text for details. (Rickett and Lumley, 1998; gure courtesy Rickett, Stanford Exploration Project, and 4th Wave
Imaging; data courtesy Chevron.)

and after cross-equalization. The bubbles correspond to


the location of the injection wells.
The 4-D seismic anomalies are characterized as differences between time-lapse 3-D data that are present
after cross-equalization as exemplied by Figure 11.55. Calibration of these anomalies often is ambiguous,
in that, they may be attributable to changes in one or
more of the reservoir conditions, such as change in uid
saturation caused by water displacing oil, pore pressure
change caused by injection, or a temperature change
caused by steam injection (Tura and Lumley, 1999).
Although signicant progress has been made in the
4-D seismic method, its value in determining dynamic
reservoir properties is just beginning to be demonstrated. The information regarding the dynamic reservoir properties much sought after by the production engineer includes changes in oil saturation, water saturation, and pore pressure. Future developments in seismically driven reservoir characterization and monitoring
should contribute signicantly to optimum management
of oil and gas elds.

11.6 4-C SEISMIC METHOD


Throughout this textbook, we discussed processing, inversion, and interpretation of compressional or P -wave
seismic data. Specically, we had in mind a compressional seismic source and the reected signal recorded
by each receiver being a compressional seismic waveeld. Recall from Section 11.2 that for non-normal incidence at a layer boundary, an incident compressional
plane wave is partitioned into not just reected and
transmitted compressional-wave components, but also
reected and transmitted shear-wave, or S-wave components. Hence, a fraction of the incident P -wave is
converted into a reected S-wave. The amplitudes of the
individual components are described by the Zoeppritz
equations (11-12). By way of prestack amplitude inversion of nonzero-oset P -wave data described in Section
11.2, we are then able to make an estimate of the Swave reectivity as an AVO attribute (equations 11-37
and 11-49).

1916

Seismic Data Analysis

FIG. 11.5-4. Time-lapse 3-D seismic data: (a) an inline section and (b) a time slice from the time image volume associated
with the 1979 survey; (c) an inline section and (d) a time slice from the time image volume associated with the 1991 survey.
The interior of the broken yellow line denes the overlapping area between the two surveys. See text for details. (Rickett and
Lumley, 1998; gure courtesy Rickett, Stanford Exploration Project, and 4th Wave Imaging; data courtesy Chevron.)

Reservoir Geophysics

1917

FIG. 11.5-5. Amplitude spectra of the two time-lapse 3-D seismic data as in Figure 11.5-4 (a) before and (b) after crossequalization; (c) a dierence section following cross-equalization. (Rickett and Lumley, 1998; gure courtesy Rickett, Stanford
Exploration Project, and 4th Wave Imaging; data courtesy Chevron.)

In a conventional marine seismic survey, we cannot record P -to-S converted-wave energy even if we deploy sensors that can register the shear-wave energy.
This is because the upcoming converted-wave energy
is not transmitted through the water column to reach
the recording cable since uids cannot support shear
strain. Thus, to capture the converted-wave energy, we
need to record it at the water bottom using an oceanbottom cable (OBC). And to record it, we need to use
geophones that register velocity of the particle motion
that is perpendicular to the direction of the wave propagation. Since the upcoming wave is primarily in the vertical direction, we need to use a geophone that records
the particle motion in the horizontal direction. In fact,
for a good reason that will be obvious later in the section, we need to deploy not one, but two horizontal geophones that are oriented perpendicular to one another.

To complement the recording of the pressure wave by


a hydrophone, again for a reason that will be obvious
later in the section, we might also wish to record the
vertical component of the particle motion using a vertical geophone. Hence, an OBC recording is done using
three geophones and one hydrophone for each receiver
unit along the cable, making it a four-component (4-C)
seismic survey. The nal product from the analysis of a
4-C survey data is a pair of P -wave and S-wave image
sections (in the case of a 2-D survey) or volumes (in
the case of a 3-D survey). Strictly the P -wave data are
associated with P -to-P reections and S-wave data are
associated with P -to-S converted waves. Heretofore, we
shall refer to these two wave types as P P and P S, respectively, so as to explicitly indicate that they both
are generated by a P -wave source. (An S-wave source
would have generated SS and SP data.)

1918

Seismic Data Analysis

Much of the P -to-S conversion takes place, not at


the water bottom, but at reectors that correspond
to layer boundaries with signicant contrast in elastic properties (Thomsen, 1998). This fortuitous phenomenon is caused by the very low speeds of shear waves
in seabed sediments (Theilen et al., 1997; Li and Yuan,
1999).
In this section, we shall briey review acquisition
and analysis techniques specic to 4-C seismic data. Before we proceed, however, we need to ask the question
why we would want to go through the expense of conducting a 4-C OBC survey. Is there any exploration
or development objective that we cannot achieve using conventional P -wave data, but we can with S-wave
data? To answer this crucial question, rst, we make
reference to one of the AVO interpretation strategies
discussed in Section 11.2. Specically, by AVO inversion
of prestack amplitudes, we wish to estimate P -wave and
S-wave reectivities and do a crossplot analysis to infer
uid types and saturation in reservoir rocks. If, instead
of indirectly extracting these AVO attributes from conventional P -wave data, we record both P P and P S data
by a 4-C survey, we may be able to broaden our understanding of properties of reservoir uids and rocks.
Known potential applications of converted-wave
data are summarized below (Caldwell, 1999; Zhu et al.,
1999; Gaiser, 1999b).
(a)
(b)
(c)
(d)
(e)
(f)
(g)
(h)
(i)

contact, on the other hand, corresponds to a signicant contrast on the P -wave velocity curve. While the
position and geometry of the oil-water contact are important in terms of the production history of the eld,
this is not the only strategic information that is needed
for the development. Specically, for production, it is
the top-reservoir boundary that needs to be delineated
accurately so as to place the horizontal well trajectory
close to the top and avoid missing a signicant vertical
oil column.
Compare the P P section derived from the conventional streamer 3-D survey and the P S section derived
from the 4-C OBC survey, both of the same vintage,
shown in Figures 11.6-1b and 11.6-1c. The P P section shows a strong event at 2 s that corresponds to
the strong contrast on the P velocity curve associated
with the oil-water contact (Figure 11.6-1a). Neverthe-

Imaging beneath gas plumes,


Imaging beneath salt domes,
Imaging beneath basalts,
Delineating reservoir boundaries with a higher Swave impedance contrast than P -wave impedance
contrast,
Dierentiating sand from shale,
Detection of uid phase change from oil-bearing to
water-bearing sands,
Detection of vertical fracture orientation,
Mapping hydrocarbon saturation, and
Mapping oil-water contact.

We now refer to a few examples illustrating the use


of P S data. Figure 11.6-1a shows portions of a dipole
sonic log measured at a well from a producing eld.
The S-wave velocity curve shows a marked contrast at
the top- (A) and base- (B) reservoir unit. It, however,
does not show a signicant contrast at oil-water contact
(OWC). The P -wave velocity curve shows a dierence
in the gradients within the postreservoir and reservoir
units, but it does not show a marked contrast at the topreservoir boundary as does the S-wave velocity curve
(A). This is because of a lack of acoustic impedance
contrast between the shales of the postreservoir unit
and the oil sands of the reservoir unit. The oil-water

FIG. 11.5-6. Dierence section after cross-equalization of


the time-lapse data as in Figure 11.5-4. (Rickett and Lumley,
1998; gure courtesy Rickett, Stanford Exploration Project,
and 4th Wave Imaging; data courtesy Chevron.)

Reservoir Geophysics
less, the top reservoir is nearly impossible to identify on
this section. The P S section, on the other hand, shows
two strong events with irregular geometry at about 3.6
s and 3.8 s. These events correspond to the strong contrasts on the S velocity curve associated with the topand base-reservoir boundaries labeled A and B in Figure
11.6-1a, respectively.
It is important to note that the P S section is not a
replacement for the P P section; instead, they are complementary. The P P section provides the information
about the oil-water contact while the P S section provides the information about the top-reservoir boundary.
Both are needed for optimum development of the eld.
Another case for converted-wave data is shown
in Figure 11.6-2. The P P section shows a zone of
gas plume associated with the underlying leaky reservoir. The gas-saturated formations cause amplitude and
traveltime distortions of the P -wave that passes through
the anomalous zone. This can make the geometry of the
underlying reservoir unit dicult to delineate. The gas
plume actually represents a complex overburden that
gives rise to strong lateral velocity variations. As such,
the complexity of the overburden can be resolved by
earth modeling in depth and the underlying reservoir
zone can, in some cases, be imaged by prestack depth
migration with an acceptable accuracy. The S-wave, on
the other hand, is relatively unscathed by the presence
of gas within the overburden; hence, the P S data provide a more accurate time image of the reservoir zone as
compared to the time image derived from the P P data.
Note also from both the sections in Figure 11.6-2 and
the 3-D image volumes shown in Figure 11.6-3 that the
reectors within the overburden above the gas plume
zone are stronger in the P P data compared to the P S
data. This is because only a fraction of the incident
P -wave energy is converted to S-wave energy at layer
boundaries. Once again, the P P and P S data complement each other, and it certainly is not one or the other.
While the P S section provides a better image of the
reservoir structure, the P P section clearly indicates the
presence of a gas plume above. This anomalous pressure
zone has to be taken seriously during well planning.

Recording of 4-C Seismic Data


Marine 4-C data are recorded by using ocean-bottom
cables with receiver units, each containing one hydrophone to detect the pressure waveeld and three geophones to detect particle motions in a Cartesian system.
The receivers used in marine multicomponent surveys
are usually of gimballed type; as such, the vertical geophone is guaranteed to measure the vertical component

1919

of the particle motion. The two horizontal components


measure the particle motions in two orthogonal directions, and they are intended to be oriented in such positions that one of them is aligned in the direction of the
receiver cable. Shown in Figure 11.6-4 is a diagram of
an ideal three-component geophone layout that would
be possible to achieve in land surveys. The orientations
of the three components coincide with a right-handed
Cartesian coordinate system. This means that the vertical z-component is positive downward, while the inline
x-component is dened to have positive direction when
the crossline y-component is clockwise with respect to
the x-component.
The geophone orientation of the layout shown in
Figure 11.6-4 is not achievable in an ocean-bottom survey. Although the vertical geophone is indeed oriented
in the vertical direction and it measures the particle motion as positive downward, the two horizontal geophones
are not guaranteed to be in the inline and crossline directions. Instead, these two geophones position themselves at various dierent, still orthogonal, directions.
As a result, the horizontal geophones associated with a
common-shot record measure particle motions in arbitrary directions, rather than the desired common inline
and crossline directions (Figure 11.6-5). This arbitrary
horizontal geophone orientation is primarily caused by
the seabed conditions such as currents, unconsolidated
sediments, and the roughness of the seabed surface.
The sensor systems used in OBC surveys are of two
types the node systems and cable systems. A node is
an individual unit that houses a single hydrophone and
three geophones in the Cartesian orientation. The nodes
are pressed into the seabed sediments by a remotely operated vehicle. Most nodal systems are derivatives of the
SUMIC system pioneered by Statoil. The more widely
used cable systems deploy dierent designs for housing
the hydrophones and geophones. The cable usually is
dragged a certain distance and then draped down to
the seabed along the desired traverse.
The recording geometry for a 4-C OBC survey is
an adaptation of a typical land 3-D survey geometry
(Section 7.1). As illustrated in Figure 11.6-5, two or
more cables are laid down on the seabed parallel to each
other, and data are recorded using a conventional seismic vessel with source locations aligned in the direction
perpendicular to the receiver lines.
Shown in Figure 11.6-6 is a composite commonshot gather from a 4-C OBC survey made up of four
records. The individual components are the hydrophone
record, and the inline, crossline and vertical geophone
records. This survey was conducted using two receiver
cables laid in parallel, 300-m apart on the seabed. While
the total cable length is 6000 m, maximum oset for

1920

Seismic Data Analysis

FIG. 11.5-7. Amplitude extractions along a time slice from two time-lapse 3-D seismic data recorded in 1996 and 1997, (a)
before and (b) after cross-equalization. (Ecker et al., 1999; gure courtesy Ecker et al. and Chevron, and 4th Wave Imaging;
data courtesy Chevron.)

some shots was up to 9000 m. Each cable carried 240


receiver units at 25-m interval; hence, each of the four
records shown in Figure 11.6-6 contains 240 traces. The
same common-shot gather is displayed in Figure 11.6-7
with AGC applied in order to better examine the signal
and noise character in each record. Note that the hydrophone and vertical geophone records exhibit events
with similar moveout since they both contain P -waves.
The records associated with the two horizontal components exhibit events with much larger moveout since
these records contain S-waves which travel with a slower

velocity than P -waves, almost twice as slow in many


rock types. Therefore, when identifying the same event
on both a P -wave record and an S-wave record, keep in
mind that the event time in the latter can be twice the
event time in the former. This means that if you plan
a 5-s record length for conventional P -wave data, you
would need to record the 4-C data using a 10-s duration. The slower S-wave velocities would also result in
a much larger moveout on the S-wave record compared
to the moveout on the P -wave record. This means that
spatial aliasing would be more serious when applying

Reservoir Geophysics

1921

FIG. 11.6-1. (a) Portions of a dipole sonic log that shows the S-wave (left) and the P -wave (right) velocity curves; (b) a
portion of a migrated P P streamer section, and (c) a portion of a migrated P S OBC section. (Courtesy MacLeod et al., 1999;
data courtesy Chevron and Schlumberger Geco-Prakla.)

1922

Seismic Data Analysis

FIG. 11.6-2. (a) A migrated P P section, and (b) a migrated P S section. (Figure courtesy Probert et al., 1999, and Schlumberger Geco-Prakla; data courtesy BP-Amoco.)

multichannel processing applications, such as f k ltering or migration, to S-wave data than P -wave data.
Figure 11.6-8 shows a common-receiver gather
which was created by sorting the common-shot gather
data as in Figure 11.6-6. The same receiver gather with
AGC applied is shown in Figure 11.6-9. The horizontal
geophone records from both the common-shot gather
(Figure 11.6-7) and the common-receiver gather (Figure 11.6-9) exhibit events with relatively more complex
moveout than those observed on the hydrophone or vertical geophone records. You should not expect a one-toone correspondence between the events on the two sets
of records. Because layer boundaries give rise to dierences in P -wave and S-wave impedance contrasts, there
may be some events that appear on both records with
dierent strengths, or some events may be present in
one record set and absent in the other.
Figure 11.6-10 shows a close-up portion of the composite shot gather shown in Figure 11.6-7 and the spectra of the individual components. Note the polarity reversal from one side of the cable to the other on the in-

line record as a consequence of the right-handed recording convention (Figure 11.6-5). The frequency content
of the hydrophone data appears to be broader than
the vertical geophone component; this is because of the
imperfect coupling of the geophones with the seabed
sediments. This dierence in bandwidth because of the
coupling issue is observed also in the common-receiver
gather shown in Figure 11.6-11.
Acquisition of 4-C OBC data is dierent from conventional streamer recording in respect of the receivers.
In fact, it is like land acquisition at the seabed. When
the seabed has irregular geometry, it gives rise to both
long- and short-wavelength statics. Therefore, in processing 4-C data, receiver statics need to be calculated
and applied to both P P and P S data.
Gaisers Coupling Analysis of Geophone Data
Variations in geophone coupling contaminate signal amplitudes registered by the geophone components, and

Reservoir Geophysics

1923

FIG. 11.6-3. 3-D image volumes associated with (a) P P data and (b) P S data as in Figure 11.6-2. (Figure courtesy Probert
et al., 1999, and Schlumberger Geco-Prakla; data courtesy BP-Amoco.)

need to be compensated for in a surface-consistent manner. Because of coupling problems, what is recorded
by each one of the three geophones is not exactly the
same as the ground motion at the seabed. A frequencydomain model equation that relates the recorded signal components {X (), Y (), Z ()} by the three geophones in the inline, crossline, and vertical directions
(x, y, z), respectively, and the actual ground motions in
the three orthogonal directions {X(), Y (), Z()} is
given by (Gaiser, 1998)

X ()
I 0
0
X()
Y () = 0 Cy Cz Y () , (11 66)
Z ()
0 Vy Vz
Z()
where is the angular frequency, I is unity, and the
nonzero elements Cy , Cz , Vy , and Vz describe the coupling response of the geophones.
Note from equation (11-66) that X () = X();
this means that we assume that the inline geophone is
perfectly coupled. Since the inline geophone is guided by
the cable itself, this is considered a valid assumption in
practice. Whereas the vertical and crossline geophones
are not coupled completely hence the nonzero elements of the coupling matrix. The imperfect coupling

leads to vertical and crossline geophone signals mutually contaminating each other in a manner that can be
modeled by equation (11-66).
We wish to estimate the ground motion vector
{X(), Y (), Z()}; this requires inverting equation
(11-66) as given by Gaiser (1998)

X()
D
0
0
X ()
1
Y () =
0
Vz Cz Y () ,
D
Z()
0 Vy Cy
Z ()
(11 67)
where D = Vz Cy Vy Cz and is the determinant of the
coupling matrix in equation (11-66).
From the matrix equation (11-67), write explicitly
the recovered ground motions
Y () =
and

Cz
Vz
Y ()
Z ()
D
D

(11 68a)

Vy
Cy
(11 68b)
Y () +
Z ().
D
D
The coupling compensation operators are estimated in a surface-consistent manner (Taner and
Koehler, 1981) with the constraint that, following the
Z() =

1924

Seismic Data Analysis

Reservoir Geophysics

1925

1926

Seismic Data Analysis

rotation, the energy of the transverse component is minimum.


Gaiser (1998) reported a coupling experiment to
verify the validity of the coupling theory described
above. Figure 11.6-12a,b,c show the inline, crossline,
and vertical geophone records obtained from an OBC
survey. In the same gure, the record associated with
the crossline geophone is shown after compenating for
coupling (Figure 11.6-12d). To study the validity of
the compensation based on equation (11-68), a diver
rmly planted the receiver unit into the seabed and the
recording was repeated. The resulting crossline record is
shown in Figure 11.6-12e. If the coupling theory holds,
then the records in Figures 11.6-12d,e should look very
similar. Dierences may be attributed to poor coupling
of the planted receiver unit.
Figure 11.6-13 shows the result of surfaceconsistent coupling analysis. To apply the coupling
corrections, scale the amplitudes in a given geophone
record by the product of the source scalar and the associated component scalar. Figure 11.6-14 shows the
common-shot gather as in Figure 11.6-6 after the application of surface-consistent amplitude corrections (Figure 11.6-12). The same shot gather with AGC is shown
in Figure 11.6-15. The common-receiver gather as in
Figure 11.6-8 after the application of surface-consistent
amplitude corrections (Figure 11.6-12) is shown in Figure 11.6-16. The same receiver gather with AGC is
shown in Figure 11.6-17. To examine the degree of compensation for dierences in geophone coupling, refer to
the close-up displays shown in Figures 11.6-18 and 11.619.
Processing of P P Data
The next step in processing of the 4-C seismic data
is to calibrate the vertical geophone component Z(t)
and sum it with the hydrophone component P (t) to
obtain the total P P data. This dual-sensor summation
is done to remove water-column reverberations (Barr
and Sanders, 1989). Calibration may involve just a single scalar applied to the entire geophone data prior to
summation with the hydrophone data. More sophisticated calibration techniques include the application of
surface-consistent scalars computed for each receiver location or application of surface-consistent scalars computed for each receiver location and for each frequency
component (Dragoset and Barr, 1994; Paenholz and
Barr, 1995; Soubaras, 1996).
The merged P P data are now ready for conventional processing. First, apply a vertical time shift that
is equal to the water depth divided by water velocity to
bring the receivers from the seabed to the same datum
as the shots. If the water depth is greater than 100 m,

the vertical shift may not be valid; instead, the datuming may have to be done by wave-equation datuming
(Section 8.1).
The remaining prestack processing sequence for the
P P data is no dierent from the land data processing sequence and includes geometric spreading correction, deconvolution, refraction and residual statics corrections
to account for the variations in the seabed geometry, velocity analysis, NMO and DMO corrections. The poststack processing sequence typically includes deconvolution, time-variant ltering, and migration. Shown in
Figure 11.6-20 is a CMP gather associated with the P P
data as in Figure 11.6-14. The CMP stack is shown in
Figure 11.6-21. The average fold of coverage is 150.
Rotation of Horizontal Geophone Components
Return to the OBC geometry shown in Figure 11.6-5. As
discussed earlier in the section, by using gimballed geophones, the vertical geophone orientation can be maintained in a true vertical direction. The two horizontal
geophones, however, cannot be oriented along the desired inline and crossline directions. Instead, they align
themselves in arbitrary orientations from one receiver
station to the next. We need to realign the horizontal
geophones associated with one common-shot gather to
a common orientation. One common orientation that
is the source-centered Cartesian coordinates is shown
in Figure 11.6-22. This means that the horizontal geophones of all receivers that contribute traces to the shot
station with a circle around it are rotated from inlinecrossline (x, y) coordinates (the acquisition coordinate
system) to radial-transverse (r, t) coordinates relative to
the source location (the processing coordinate system).
As a result, the radial geophone response will represent
the horizontal particle motion in the source-receiver
plane and the transverse geophone response will represent the horizontal particle motion perpendicular to the
radial response. Following the rotation, a common-shot
or a common-receiver gather associated with the radial
component will comprise traces with radial responses in
the source-receiver azimuthal directions.
Figure 11.6-23 shows a common-receiver gather associated with the inline and crossline geophone components before and after rotation. The equations for coordinate transformation of the particle motions from
inline-crossline (x, y) ccordinates to radial-transverse
(r, t) coordinates are (Li and Yuan, 1999)

R(t)
cos
sin
X(t)

, (11 69)
T (t)
sin cos
Y (t)
(text continues on p. 1933)

Reservoir Geophysics

1927

FIG. 11.6-6. Individual components of a common-shot gather from a 4-C survey, (a) the hydrophone, (b) inline, (c) crossline,
and (d) vertical geophone components. (Field data related to Figures 11.6-6 through 11.6-29 are courtesy Chevron.)

1928

Seismic Data Analysis

FIG. 11.6-7. Individual components of a common-shot gather as in Figure 11.6-6, but displayed with AGC, (a) the hydrophone, (b) inline, (c) crossline, and (d) vertical geophone components.

Reservoir Geophysics

1929

FIG. 11.6-8. Individual components of a common-receiver gather from a 4-C survey, (a) the hydrophone, (b) inline, (c)
crossline, and (d) vertical geophone components.

1930

Seismic Data Analysis

FIG. 11.6-9. Individual components of a common-receiver gather as in Figure 11.6-8, but displayed with AGC, (a) the
hydrophone, (b) inline, (c) crossline, and (d) vertical geophone components.

Reservoir Geophysics

1931

FIG. 11.6-10. A close-up portion of the composite common-shot gather shown in Figure 11.6-6 and the spectra of the
individual components, (a) the hydrophone, (b) inline, (c) crossline, and (d) vertical geophone components.

1932

Seismic Data Analysis

FIG. 11.6-11. A close-up portion of the composite common-receiver gather shown in Figure 11.6-8 and the spectra of the
individual components, (a) the hydrophone, (b) inline, (c) crossline, and (d) vertical geophone components.

Reservoir Geophysics
where the rotation angle is as labeled in Figure 11.623c, Y (t) and X(t) are the inline and crossline components as recorded in the eld following the compensation for variations in coupling, and R(t) and T (t) are the
rotational and transverse components after rotation.
It is generally assumed that most signicant signal components reections, diractions, and converted waves, are polarized in the source-receiver direction (Gaiser, 1998). This means that, following the rotation of the horizontal geophone components, the radial
component would contain most of the P S energy while
the transverse component would contain negligible P S
energy (Figure 11.6-23). If the transverse component
does contain anomalously high level of signal energy,
it may be attributable to anisotropy that causes shearwave splitting in fractured rocks. We shall review this
pheonomenon briey in Section 11.7.
Figure 11.6-24 shows the common-shot gather as
in Figure 11.6-14 after rotation of the horizontal components. The hydrophone record (rst record from left)
and the vertical geophone record (fourth record from
left) are the same in both gures. The second and third
record from the left in Figure 11.6-14 represent the inline and crossline geophone components of the particle
motion; whereas, the second and third record from the
left in Figure 11.6-24 represent the radial and transverse
geophone components of the particle motion. Note from
the AGC applied version of the same shot gather in
Figure 11.6-25 that the radial component, unlike the
inline component shown in Figure 11.6-15, does not
exhibit the polarity reversal at zero oset. Also, the
transverse component in Figure 11.6-25 contains relatively weak signal energy when compared to the radial
component. These observations are also veried in the
common-receiver gather shown in Figure 11.6-26. Compare the AGC applied version of the receiver gather before (Figure 11.6-16) and after (Figure 11.6-27) rotation
and note the switch in polarity and low signal level in
the transverse component. The spectral analysis of the
shot and receiver gathers are shown in Figures 11.6-28
and 11.6-29; compare these results with Figures 11.6-18
and 11.6-19.

Common-Conversion-Point Binning
We learned in Section 11.2 that an incident P -wave
is partitioned at a layer boundary into reected and
transmitted P - and S-wave components. Consider the
raypath geometry in Figure 11.6-30a for an incident P wave generated by the source S1 and a at reector.
The reection angle for the P P -wave is equal to the
angle of incidence; however, the reection angle for the

1933

P S-wave is smaller than the angle of incidence. As a


result, the P P reection will follow a symmetric raypath and be recorded at receiver location R2 , while the
P S reection will follow an asymmetric raypath and be
recorded at receiver location R1 .
Now consider the common-midoint (CMP) raypath
geometry for a source-receiver pair S1 R1 shown in
Figure 11.6-30b. There are two reection arrivals at the
receiver location R1 associated with the P P and P S
raypaths. The reection point B at which the incident
P -wave is converted to the S-wave is displaced in the
lateral direction by some distance d away from the reection point A at which the incident P -wave is reected and recorded at the same receiver location R1
as the converted S-wave. This means that, for an earth
model with at layers, the P P -wave reection points
coincide with the midpoint location (Figure 11.6-31a);
whereas, the P S conversion points do not (Figure 11.631b). As a direct consequence of this observation, the
notion of a CMP gather based on sorting P P data from
acquisition coordinates source and receiver, to processing coordinates midpoint and oset, such that
traces in the gather have the same midpoint coordinate,
is not applicable to P S data. Instead P S data need to
be sorted into common-conversion-point (CCP) gathers
such that traces in this gather have the same conversion
point coordinate.
An important aspect of CCP sorting is that the
asymmetric raypath associated with the P S reection
gives rise to a periodic variation in fold of the CCP gathers. As for the conventional P -wave data with variations
in fold caused by irregular recording geometry, amplitudes of the stacked P S data are adversely aected by
the variation in the CCP fold (Eaton and Lawton, 1992;
Li and Yuan, 1999). Just as one resorts to exible bin
size in the processing of 3-D seismic data to accommodate variations in fold, the same strategy may be
applied for the P S data processing.
Binning the P S data into CCP gathers requires
knowledge of the conversion-point coordinate xP . Referring to Figure 11.6-31b, note that the conversion-point
coordinate follows a trajectory indicated by the broken
curve that, in general, depends on the reector depth
(Tessmer and Behle, 1988).
To derive an expression for xP , refer to the geometry of the P S-raypath shown in Figure 11.6-32. By
Snells law, we know that
sin 0
sin 1
=
,

(11 70)

where and are the P -wave and S-wave velocities,


respectively, and 0 is the P -wave angle of incidence
and 1 is the reection angle for the converted S-wave.
(text continues on p. 1946)

1934

Seismic Data Analysis

Reservoir Geophysics

1935

1936

Seismic Data Analysis

FIG. 11.6-14. The composite common-shot gather as in Figure 11.6-6 after coupling corrections of Figure 11.6-13, (a) the
hydrophone, (b) inline, (c) crossline and (d) vertical geophone components (Data courtesy Chevron.)

Reservoir Geophysics

1937

FIG. 11.6-15. The composite common-shot gather as in Figure 11.6-14, but displayed with with AGC, (a) the hydrophone,
(b) inline, (c) crossline, and (d) vertical geophone components.

1938

Seismic Data Analysis

FIG. 11.6-16. The composite common-receiver gather as in Figure 11.6-8 after coupling corrections of Figure 11.6-13, (a)
the hydrophone, (b) inline, (c) crossline, and (d) vertical geophone components.

Reservoir Geophysics

1939

FIG. 11.6-17. The composite common-receiver gather as in Figure 11.6-16, but displayed with with AGC, (a) the hydrophone,
(b) inline, (c) crossline, and (d) vertical geophone components.

1940

Seismic Data Analysis

FIG. 11.6-18. A close-up view of the composite common-shot gather after coupling corrections as in Figure 11.6-14, and the
spectra of the individual components, (a) the hydrophone, (b) inline, (c) crossline, and (d) vertical geophone components.

Reservoir Geophysics

1941

FIG. 11.6-19. A close-up view of the composite of the common-receiver gather after coupling corrections as in Figure 11.616, and the spectra of the individual components, (a) the hydrophone, (b) inline, (c) crossline, and (d) vertical geophone
components.

1942

Seismic Data Analysis

Reservoir Geophysics

1943

FIG. 11.6-21. The CMP stack derived from the P P -data as in Figure 11.6-14. (Processing by Orhan Yilmaz, 1999).

1944

Seismic Data Analysis

Reservoir Geophysics

1945

1946

Seismic Data Analysis

From the geometry of Figure 11.6-32, note that


xP
(11 71a)
sin 0 =
2
xP + z 2
and
sin 1 =

xS
x2S + z 2

(11 71b)

where xP and xS are the lateral distances from the


CCP to the source and receiver locations, respectively.
Substitute equations (11-71a,b) into equation (11-70),
square and rearrange the terms to get
x2S
2 x2 + z 2
= 2 2S
.
2
xP
xP + z 2

(11 72a)

Apply some algebraic manipulation to solve equation


(11-72a) for xS
xS =

xP
x2
2 + 2 1 P2
z

(11 72b)

where = /. Finally, substitute the relation xS =


x xP , where x is the source-receiver oset, into equation (11-72b) to get the desired expression
x2
2 + 2 1 P2
z
x.
(11 72c)
xP =
x2P
2
2
1+ + 1 2
z
From Figure 11.6-31b, note that the CCP location
moves closer to CMP location as the depth of the reector increases. At innite depth, the CCP location
reaches an asymptotic conversion point (ACP) (Fromm
et al., 1985). In the limit z , equation (11-72c)
gives the ACP coordinate xP with respect to the source
location

x.
(11 73a)
xP =
1+
Since < , the conversion point is closer to the receiver location than the source location (Figure 11.631b). The displacement d = xP x/2 of the asymptotic
conversion point from the midpoint is, by way of equation (11-73a),
1
d=
2

1
+1

x.

(11 73b)

While CCP binning may be performed using the


ACP coordinate given by equation (11-73a), more
accurate binning techniques account for the depthdependence of the CCP coordinate xP based on a solution to equation (11-72c) (Tessmer and Behle, 1988;
Zhang and Robinson, 1992). Because of the quartic form
of equation (11-72c) in terms of xP , an iterative solution may be preferred in practice (Zhang and Robinson,

1992; Zhang, 1996; Yuan and Li, 1997). The iteration


may be started by substituting the asymptotic form of
xP given by equation (11-73a) into the right-hand side
of equation (11-72c). The new value of xP may then
be back substituted into equation (11-72c) to continue
with the iteration.
Whatever the estimation procedure, note from
equation (11-72c) that xP depends both on depth to
the reector and the velocity ratio = /. Unless a
value for the velocity ratio is assumed, it follows that
CCP binning requires velocity analysis of P S data to
determine the velocity ratio . Additionally, an accurate
CCP binning strictly requires the knowledge of reector
depths; thus, the advocation of an implicit requirement
that 4-C seismic data analysis should be done in the
depth domain. This requirement may be waivered if we
only consider a horizontally layered earth model as in
the next subsection.

Velocity Analysis of PS Data


From the raypath geometry of Figure 11.6-32, it follows
that
1
1
2
(11 74)
x2P + z 2 +
x xP + z 2 ,
t=

where t is the two-way P S reection traveltime from


the source to the conversion point to the receiver, and
z is the reector depth.
Set x = xP = 0 in equation (11-74) to see that the
two-way zero-oset P S traveltime is given by
t0 =

1
1
+
z.

(11 75)

Substitute equation (11-75) into equation (11-74) for


the depth variable z to get
t=

x2P +

2
+1

x xP

2
2 t0

2
+1

2
2 t0 ,

(11 76)

where = /.
Equation (11-76) describes the P S-wave moveout
observed on a CCP gather. Although it is derived for
a single at layer in a constant-velocity medium, this
equation also is applicable to a horizontally layered
earth model. In that case, and refer to the P - and
S-wave rms velocities.
Note from equation (11-76) that the asymmetric
raypath associated with the P S reection shown in
(text continues on p. 1955)

Reservoir Geophysics

1947

FIG. 11.6-24. The composite common-shot gather as in Figure 11.6-14 after rotation of inline and crossline components to
radial and transverse components, (a) the hydrophone, (b) radial, (c) transverse, and (d) vertical geophone components.

1948

Seismic Data Analysis

FIG. 11.6-25. The composite common-shot gather as in Figure 11.6-24, but displayed with AGC, (a) the hydrophone, (b)
radial, (c) transverse, and (d) vertical geophone components.

Reservoir Geophysics

1949

FIG. 11.6-26. The composite common-receiver gather as in Figure 11.6-16 after rotation of inline and crossline components
to radial and transverse components, (a) the hydrophone, (b) radial, (c) transverse, and (d) vertical geophone components.

1950

Seismic Data Analysis

FIG. 11.6-27. The composite common-receiver gather as in Figure 11.6-26, but displayed with AGC, (a) the hydrophone,
(b) radial, (c) transverse, and (d) vertical geophone components.

Reservoir Geophysics

1951

FIG. 11.6-28. A close-up view of the composite common-shot gather after rotation as in Figure 11.6-24, and the spectra of
the individual components, (a) the hydrophone, (b) radial, (c) transverse, and (d) vertical geophone components.

1952

Seismic Data Analysis

FIG. 11.6-29. A close-up view of the composite common-shot gather after rotation as in Figure 11.6-26, and the spectra of
the individual components, (a) the hydrophone, (b) radial, (c) transverse, and (d) vertical geophone components.

Reservoir Geophysics

1953

FIG. 11.6-30. (a) Conversion of an incident P -wave to a reecting S-wave; (b) P -to-P and P -to-S reection at a at
reecting interface. The arrows along the raypaths represent the direction of particle motion. See text for details.

1954

Seismic Data Analysis

FIG. 11.6-31. Geometry of (a) common-midpoint (CMP) raypaths, and (b) common-conversion-point (CCP) and raypaths.
At innite depth, the CCP location reaches an asymptotic conversion point (ACP). See text for details.

Reservoir Geophysics

1955

FIG. 11.6-32. Geometry of a common-conversion-point (CCP) raypath used to derive the reection traveltime equation
(11-74) for the P S-wave.

Figure 11.6-32 gives rise to a nonhyperbolic moveout


even in the case of a at reector in a constant-velocity
medium. A way to avoid dealing with nonhyperbolic
moveout would be to make the small-spread approximation and consider the best-t hyperbola
t = t20 +

x2
2
vN
MO

(11 77)

to the traveltime trajectory associated with a P S reection on a CCP gather (Tessmer and Behle, 1988). In
equation (11-77), t and t0 mean the same traveltimes
as in equation (11-76), and vN M O is the moveout velocity for P S-wave derived from the best-t hyperbola; as
such, it is neither the P -wave velocity nor the S-wave
velocity . In fact, the Taylor expansion
of equation

(11-74) yields the relation vN M O = (Fromm et


al., 1985). By assuming a hyperbolic moveout described
by equation (11-77), the P S data can be corrected for
moveout and stacked in the same manner as for the P P
data.
Practical experience, however, points to the unavoidable fact that the P S data exhibit strong nonhyperbolic moveout behavior. Shown in Figure 11.6-33 is a
CCP gather after hyperbolic moveout correction using
equation (11-77) and nonhyperbolic moveout correction
using equation (11-76). Note the overcorrection at far
osets within 0-2.5 s.

The need for nonhyperbolic moveout correction for


the P S data makes it compelling to conduct a multiparameter velocity analysis. Unlike velocity analysis using the hyperbolic moveout equation (11-77), where we
only need to scan for one parameter, vN M O , velocity
analysis using the nonhyperbolic moveout equation (1176) suggests scanning for three parameters the P P wave velocity , the velocity ratio = /, and the
CCP displacement xP . But, in practice, we do not have
to scan for all three parameters. Instead, the iterative
procedure described below may be followed.
(a) To begin with, note that the P P -wave velocity
can be estimated directly by velocity analysis of
the P P data set itself.
(b) We may assume an initial value for the velocity
ratio = / and estimate an intial value for xP
using equation (11-73a).
(c) Knowing and xP , use the nonhyperbolic moveout
equation (11-76) to scan for as a function of t0 .
Figure 11.6-34 shows a -spectrum computed from
the CCP gather in Figure 11.6-33a.
(d) Pick a function (t0 ) at each CCP analysis location
along the line over the survey area and derive a
(x, t0 )-section as shown in Figure 11.6-35.
(e) Use the (x, t0 )-section and the P P -wave velocity

1956

Seismic Data Analysis

FIG. 11.6-33. (a) A common-conversion-point (CCP) gather, (b) after hyperbolic moveout correction using equation (11-77),
and (c) after nonhyperbolic movoeut correction using equation (11-76). (Figure courtesy Li and Yuan, 1999.)

to calculate an updated value for xP (t0 ) from


equation (11-72c).
(f) Substitute the updated xP (t0 ) and the estimated
and into equation (11-76) to perform the nonhyperbolic moveout correction (Figure 11.6-33c).

Another strategy for velocity analysis of the P Swave data is the direct estimation of the P S-wave velocity , rather than estimating the velocity ratio .
Return to equation (11-76) and rewrite it explicitly in
terms of and as

Reservoir Geophysics

1957

FIG. 11.6-34. Analysis of CCP gather in Figure 11.6-33a for the velocity ratio, = /. (Figure courtesy Li and Yuan,
1999.)

1958

Seismic Data Analysis

FIG. 11.6-35. The velocity-ratio section derived from the analysis as in Figure 11.6-34. (Figure courtesy Li and Yuan, 1999.)

Reservoir Geophysics
t=

x2P +

2 2
+

x xP

2
2 t0

2 2
+

2
2 t0 .

(11 78)

By using the nonhyperbolic equation (11-78), follow the alternative procedure for P S-wave velocity
analysis outlined below.
(a) Again, estimate the P P -wave velocity as before
using the P P data set itself.
(b) Also, assume an initial value for = / and estimate an initial value for xP using equation (1173a).
(c) Knowing and xP , use the nonhyperbolic moveout
equation (11-78) to scan for as a function of t0 .
Figure 11.6-36 shows a CCP gather and the computed -spectrum. Compare with the -spectrum
in Figure 11.6-20 and note the dierence in the velocity ranges in the two spectra.
(d) Pick a function (t0 ) at each CCP analysis location
along the line over the survey area and derive a
(x, t0 )-section.
(e) Use the (x, t0 )-section and the P P -wave velocity
to calculate an updated value for xP (t0 ) from
equation (11-72c).
(f) Substitute the updated xP (t0 ) and the estimated
and into equation (11-78) to perform the nonhyperbolic moveout correction.
Figure 11.6-37 shows the P S-wave stack based on
the alternative procedure described above. Note the differences between this section and the P P -wave stack
shown in Figure 11.6-21. Specically, the P S-wave stack
shows some interesting reector geometry between 4-5
s; this behavior is absent within the equivalent time
window (2-2.5 s) in the P P -wave stack.
One subtle issue is related to the time at which is
specied in equations (11-76) and (11-78). The two-way
zero-oset time t0 in these equations is associated with
the P S-wave; whereas, is estimated at two-way zerooset time associated with the P P -wave. To distinguish
the two zero-oset times, rst, rewrite equation (11-75)
(P S)
for the P S two-way time t0
(P S)

t0

1
1
+
z.

(11 79a)
(P P )

For the same depth z, the P P two-way time t0


given by
(P P )

t0

2
z.

is

(11 79b)

1959

Now eliminate z between the two equations to get the


relation between the P P and P S zero-oset times
2
(P S)
(P P )
t
,
(11 80a)
=
t0
+ 0
or in terms of = /
(P P )

t0

2
(P S)
t
.
+1 0

(11 80b)
(P P )

So, in equation (11-78) is specied at time t0


given
by equation (11-80a), and in equation (11-76) is spec(P P )
ied at time t0
given by equation (11-80b).
Dip-Moveout Correction of P S Data
In Section 5.1 we learned that the DMO impulse response is in the form of an ellipse (Figure 5.1-12).
We also learned that the aperture of the DMO operator depends on the reector dip, the source-receiver
separation, the reection time of moveout-corrected
common-oset data, and the velocity above the reector. Since shear-wave velocities are generally lower than
compressional-wave velocities, given that all other factors are the same, the DMO impulse response associated
with the P S data will be dierent in shape as compared
to the DMO impulse response associated with the conventional P P data shown in Figure 5.1-12. Specically,
the slower the velocity the more the action of the DMO
operator (Section 5.1). This means that the symmetric
form of the DMO ellipse associated with the P P data
is replaced with an asymmetric form as shown in Figure 11.6-38 (Alfaraj, 1993; Alfaraj and Larner, 1993;
den Rooijen, 1991). Additionally, the resulting curve is
shifted laterally to the CCP location indicated by the
dotted trajectory in Figure 11.6-38b (Alfaraj, 1993).
The P P DMO impulse response for variable velocity v(z) increasing with depth can be formed by squeezing the P P DMO impulse response for constant velocity (Hale and Artley, 1992). Similarly, the asymmetric
shape of the P S DMO impulse response shown in Figure 11.6-38b may be formed by squeezing the P P DMO
impulse response on the source side and stretching it on
the receiver side (Alfaraj, 1993).
Another important dierence between the DMO
corrections of the P P and P S data is in respect to
the reection-point dispersal (Section E.1). We learned,
again in Section 5.1, that the DMO correction removes
the reection-point dispersal along a dipping reector.
With the P P data, reection-point dispersal is an issue only in the case of a dipping reector. However,
with the P S data, reection-point dispersal takes place
even for the case of a at reector. This is illustrated
in Figure 11.6-39 where the raypaths of Figure 11.6-32

1960

Seismic Data Analysis

Reservoir Geophysics
have been adapted to the CMP geometry. Note that you
need to apply DMO correction to the P S data even in
the absence of dip to remove the reection-point dispersal (Alfaraj, 1993). Expressed dierently, implicit to
the P S DMO correction is CCP binning that involves
mapping of amplitudes along the dotted trajectory in
Figure 11.6-38b.
Migration of PS Data
We learned in Section 4.1 that migration of the stacked
P P data can be performed by summation of amplitudes
(equation 4-5) along a hyperbolic traveltime trajectory (equation 4-4) associated with a coincident sourcereceiver pair at the surface and a point diractor at the
subsurface. The amplitude from the resulting summation is placed at the apex time of the diraction hyperbola. Prior to summation (equation 4-5), amplitude and
phase factors inferred by the Kirchho integral solution
to the scalar wave equation are applied to the stacked
data (Section 4.1).
Similarly, migration of the stacked P S data can
be conceptualized as a summation along a traveltime
trajectory associated with a coincident source-receiver
pair at the surface and a point diractor at the subsurface. The dierence between the zero-oset P P and the
P S diraction summation trajectories is in the velocities. Set x = 0 in equation (11-76) to get the traveltime
equation for the zero-oset P S diraction summation
trajectory as
+1 2 2
xP ,
(11 81)

where t is the time associated with the unmigrated P Sstack, and t0 is the time associated with the migrated
P S-stack given by equation (11-75).
Equation (11-81) is of the same form as equation
(4-4) that describes the zero-oset P P diraction summation trajectory. In fact, setting = 1 in equation
(11-81) yields equation (4-4) for the case of constant
velocity. Note from equation (11-81) that the migration
velocity for the P S data is given by /( + 1), which is
neither the P -wave velocity nor the S-wave velocity
. The P S-velocity /( + 1) can be obtained from the
P P -velocity and the velocity ratio , both estimated
from the velocity analysis described earlier in this section. Once the velocity eld is estimated, the P S-stack
can, in principle, be migrated using any of the algorithms described in Chapter 4.
t2 = t20 +

11.7 SEISMIC ANISOTROPY


A medium is anisotropic if its intrinsic elastic properties, measured at the same location, change with direction, and it is isotropic if its properties do not change

1961

with direction (Winterstein, 1990). Most of seismic data


analysis is based on the assumption that the subsurface
behaves seismically isotropic. A case of anisotropy that
is of interest in exploration seismology is the change in
velocity with direction.
A medium is tranversely isotropic if its elastic properties do not change in any direction perpendicular to
an axis of symmetry. The usual meaning of seismic
anisotropy is variation of seismic velocity, which itself
depends on the elastic properties of the medium, with
the direction in which it is measured (Sheri, 1991).
Anisotropic variation of seismic velocity must not be
confused with the source-receiver azimuthal variation of
moveout velocity for a dipping reector in an isotropic
medium (equation 7-3).
There are two cases of seismic anisotropy that we
shall review in this section, both of which are special
cases of transverse isotropy. For convenience, consider
a horizontally layered earth model. First, is the vertical transverse isotropy, otherwise simply referred to
as transverse isotropy, for which velocities do not vary
from one lateral direction to another, but vary from one
direction to another on a vertical plane that coincides
with a given lateral direction. Horizontal bedding and
fracturing parallel to the bedding produce transverse
isotropy. Second, is the horizontal transverse isotropy,
otherwise known as azimuthal anisotropy, for which velocities vary from one lateral direction to another. Fracturing in a direction other than the bedding direction
gives rise to azimuthal aniostropy.
Transverse isotropy stems from the fact that, as a
result of a depositional process, velocities in a layer are
dierent in the direction of bedding and the direction
perpendicular to the bedding. Specically, ne layering
of isotropic beds that constitute the depositional unit
give rise to an anisotropy with its axis of symmetry perpendicular to the bedding plane. Azimuthal anisotropy
can stem from the fact that, as a result of a tectonic process, rock material associated with a layer may have different rigidity in dierent azimuthal directions. A specic example is secondary fracturing in rocks whereby
the velocity in the fracture direction is higher than the
velocity in the orthogonal direction, again, giving rise
to an anisotropy with its axis of symmetry parallel to
the bedding plane.
Refer to Figure 11.7-1 to review the physical aspects of wave propagation in an anisotropic medium.
The directional change of the velocity is illustrated by
the skewed ellipse in Figure 11.7-1a, with the fast velocity in the direction of the major axis and the slow
velocity in the direction of the minor axis. Now consider

1962

Seismic Data Analysis

FIG. 11.6-37. The CCP stack derived from the P S data as in Figure 11.6-14. Compare with the CMP stack derived from
the P P data shown in Figure 11.6-21. (Processing by Orhan Yilmaz, 1999).

Reservoir Geophysics

1963

FIG. 11.6-38. Impulse responses of a DMO operator for (a) P P data, and (b) P S data (adapted from Alfaraj, 1993) for a
given source (S) and receiver (R) separation. The horizontal axis is the midpoint axis and the vertical axis is the event time
after NMO correction. The dotted curve represents the common-conversion-point (CCP) trajectory as in Figure 11.6-31b.

Huygens secondary sources situated along the wavefront A, say at time t. This wavefront actually coincides with an exploding reector dipping at an angle .
Because of the anisotropic behavior of the propagation
medium, the Huygens sources do not emanate semicircular wavefronts; instead, the wavefronts are skewed in
the direction of the fast velocity. These skewed wavefronts will form the plane wavefront B at a later time
t + t. While the energy was transmitted along the raypath SP at group velocity, the wavefront that represents
a constant phase actually traveled from position A to
B along T P normal to the wavefront at phase velocity. Because the group velocity is associated with the
raypath, it is sometimes referred to as the ray velocity. Similarly, because the phase velocity is associated
with the wavefront, it is sometimes referred to as the
wavefront velocity. Note that the wavefront angle associated with the phase velocity is dierent from the
ray angle associated with the group velocity. Only if
the medium were isotropic, Huygens secondary sources
would produce semicircular wavefronts and the phase
angle would coincide with the ray angle .

Note from Figure 11.7-1 that the zero-oset raypath SP does not make a right angle with the reector;
thus, in the case of anisotropy, the zero-oset ray is not a
normal-incident ray as would be in the case of isotropy.
This behavior can be better explained by sketching the
isotropic and anisotropic wavefronts as shown in Figure
11.7-1b. Specically, in the case of isotropic medium,
the wavefront emanating from a point P is circular and
the zero-oset ray is normal-incident to the reector.
Whereas, in the case of an anisotropic medium, the
wavefront is skewed and the zero-oset ray impinges
on the reector at a non-normal incidence angle.
An example of how velocity changes with direction in an anisotropic medium is shown in Figure 11.7-2
(Uren et al., 1990a). Based on velocity characterisitcs of
the Green River shale (Thomsen, 1986), this polar plot
of group velocities shows that, as for most rock types,
the P -wave velocity is nonelliptical, whereas the SHwave velocity behaves elliptically anisotropic. Also note
that the horizontal P -wave velocity is greater than the
vertical P -wave velocity.
We shall associate seismic aniostropy primarily with seismic velocities. Aside from anisotropic

1964

Seismic Data Analysis

FIG. 11.6-39. (a) Common-midpoint (CMP) raypath geometry for P P -data, and (b) common-conversion-point (CCP)
raypath geometry for P S-data. (Adapted from den Rooijen, 1991.)

Reservoir Geophysics
velocity analysis, this means that we will need to review those processes that are intimately inuenced by
velocity anisotropy, such as dip-moveout correction, migration and AVO analysis.
The generalized form of Hookes law, which is the
foundation of linear elastic theory, states that each
stress component can be expressed as a linear combination of all the strain components (Ocer, 1958).
Hookes law is based on the assumption that elastic deformations in solids are innitesimally small. The stiness matrix, otherwise known as the elastic modulus
matrix (Thomsen, 1986), that relates the stress components to the strain components is (Section L.1)

c11 c12 c13 c14 c15 c16


c21 c22 c23 c24 c25 c26

c
c
c
c
c
c
{cij } = 31 32 33 34 35 36 .
c41 c42 c43 c44 c45 c46

c51 c52 c53 c54 c55 c56


c61 c62 c63 c64 c65 c66
(11 82a)
The elements of the stiness matrix are the elastic constants of an elastic solid. Since this matrix is symmetric, cij = cji , there are 21 independent constants for an
elastic medium.
For an isotropic solid, the elastic behavior of which
is independent of direction at a point within the solid,
the number of independent elastic constants is only two,
known as Lames constants, and . Consequently, the
stiness matrix given by equation (11-82a) reduces to
the special form

+2

0
0
0
+2

0
0
0

+2
0
0
0

{cij }=
.
0
0
2
0
0
0

0
0
0
0
2
0
0
0
0
0
0
2
(11 82b)
To describe the P - and S-wave propagation in
isotropic solids, only two elastic parameters are needed
Lames constants, and . Other elastic parameters Youngs modulus E, Poissons ratio and bulk
modulus , and the P - and S-wave velocities can all
be expressed in terms of and (Figure 11.0-1 and
Appendix L.1).
For a transversely isotropic solid, the elastic behavior of which is the same in two orthogonal directions but
dierent in the third direction, the number of independent constants is ve (Ocer, 1958; Thomsen, 1986;
Sheri, 1991). For the more specic case of vertically
transverse isotropy (VTI), which has a vertical symmetry axis, the ve independent elastic constants are c11 ,
c13 , c33 , c44 , and c66 (Thomsen, 1986).

1965

To explicitly describe the eect of anisotropy in


wave propagation, Thomsen (1986) has elegantly redened the ve elastic constants for the VTI media
the vertical P - and S-wave velocities, 0 and 0 , in the
vertical direction,
0 =

c33
,

(11 83a)

0 =

c44
,

(11 83b)

and three constants that describe the degree of


anisotropy, , , and , in terms of the ve constants
c11 , c13 , c33 , c44 , and c66
=
=

c11 c33
,
2c33

(11 83c)

(c13 + c44 )2 (c33 c44 )2


,
2c33 (c33 c44 )

(11 83d)

c66 c44
.
2c44

(11 83e)

and
=

For most sedimentary rocks, the parameters , ,


and are of the same order of magnitude and usually
much less than 0.2 (Thomsen, 1986). In fact, the Thomsen parameters given by equations (11-83c,d,e) relate to
the case of weak anisotropy described by small values
( 1) of , , and . While applications of anisotropy
in seismic data analysis are primarily based on the assumption of weak anisotropy, these parameters still are
useful to describe the general case of transverse isotropy.
The special case of = is known as elliptical
anisotropy (Daley and Hron, 1979). The ellipticity is
associated with the shape of the wavefront expanding
from a point source. Albeit its underlying theory is simpler than the general theory of anisotropy, elliptical
anisotropy occurs in nature only rarely. Figure 11.7-3
shows a crossplot of the two Thomsen parameters,
and , for various types of sedimentary and crystalline
rocks based on eld and laboratory studies (Thomsen,
1986). Note that a few of the rock samples closely satisfy the ellipticity condition, = . Also, for most rock
types, the Thomsen parameters, and , are positive
constants less than 0.2; thus, the supporting evidence
for weak anisotropy theory.
Anisotropic Velocity Analysis
We shall conne the discussion on anisotropy primarily
to the practical case of P -wave propagation in weakly
anisotropic rocks. Consider the wavefront geometry

1966

Seismic Data Analysis

shown in Figure 11.7-1. The P -wave phase velocity is


given by (Thomsen, 1986)
() = 0 1 + sin2 cos2 + sin4 .

(11 84)

Note that the P -wave velocity depends on the


anisotropy parameters and , but not on the parameter . In fact, the SV -wave velocity also depends only
on and , while the SH-wave velocity depends only
on .
In the special case of vertical incidence, = 0,
equation (11-84) gives the vertical P -wave velocity, 0 .
In the special case of horizontal incidence, = 90 degrees, equation (11-84) gives
h = 0 (1 + ).

(11 85)

Solving for , we obtain


=

h 0
.
0

(11 86)

This equation provides a physical meaning for the


Thomsen parameter . Specically, the parameter indicates the degree of aniostropic behavior of the rock,
measured as the fractional dierence between vertical
P -wave velocity 0 and the horizontal P -wave velocity
h . Since for most rocks > 0, note that the horizontal P -wave velocity is greater than the vertical P -wave
velocity.
The normal-moveout velocity vN M O ( = 0), where
is the dip angle, for a at reector in an anisotropic
medium is given by Thomsen (1986)

(11 87)
vN M O (0) = 0 1 + 2.
In the special case of = 0, the moveout velocity is the
same as the velocity of an isotropic medium represented
by 0 (Section 3.1).
The P -wave traveltime equation for a at reector in an anisotropic medium is given by Tsvankin and
Thomsen (1994)
A4 x4
,
(11 88)
1 + Ax2
where t is the two-way time from source to reector
to receiver, t0 is the two-way zero-oset time, x is the
source-receiver oset, and A2 and A4 are coecients
which are written below to acknowledge their complexity
t2 = t20 + A2 x2 +

A2 =
A4 =

02 (1

1
,
+ 2)

(11-89a)

2( ) 1 + 2 1 02 /02 )1
t20 04 1 + 2

(11-89b)

and
A=

A4
.
A2

h2

(11 89c)

Just for comparison, we write the traveltime equation (3-11) for a at reector in an isotropic medium
using the current notation
t2 = t20 +

x2
2
vN
MO

(11 90)

For isotropic velocity analysis using equation (1190), we only need to scan one parameter the velocity
vN M O (Section 3.2). For anisotropic velocity analysis
using equation (11-88), note that we have to do a multiparameter scan involving 0 , h , 0 , , and an
impractical and unacceptable proposal.
As a way to circumvent this insurmountable problem of a multiparameter scan in velocity analysis,
Alkhalifah and Tsvankin (1995) dene a new eective
anisotropy parameter
=

.
1 + 2

(11 91)

By way of equations (11-85), (11-87), (11-91), and (1189a,b,c), and neglecting the eect of 0 , equation (1188) takes a form that is suitable for practical implementation (Alkhalifah and Tsvankin, 1995)
t2 = t20 +

x2
2
vN
MO

2
vN
MO

2x4
,
2
2
t20 vN
M O + 1 + 2 x
(11 92)

where vN M O is the at-reector moveout velocity.


Equation (11-92) indicates that the traveltime for
a reector in an anisotropic medium follows a nonhyperbolic trajectory. By setting = 0, equation (1192) reduces to the isotropic case of equation (11-90).
Note that the eect of anisotropy on reection traveltimes is most signicant at far osets. The derivation
of the anisotropic moveout equation (11-92) is based on
a single at layer. Nevertheless, in practice, it also is
applicable to the case of a horizontally layered earth
model with vertical transverse isotropy (Grechka and
Tsvankin, 1997).
Figure 11.7-4a shows CMP raypaths associated
with a at reector in a transversely isotropic medium
with a velocity behavior as shown in Figure 11.72. Transverse isotropy causes fundamental departures
from the CMP raypath geometry associated with an
isotropic medium:
(a) The zero-oset raypath is non-normal incident.
(b) A single common reection point does not exist;
instead, anisotropy has given rise to reection point
dispersal even for a at reector.
(c) The anisotropic moveout given by equation (11-92)
in general is nonhyperbolic.

Reservoir Geophysics

1967

rected events between the two-way zero-oset times


of 1.5-2 s and some undercorrected events above the
two-way zero-oset time of 1.5 s. The undercorrected
events are clearly multiples which produce distinct semblance peaks on the velocity spectrum. The overcorrected events result from one of the three possibilities
that the overcorrection is caused by the dip eect
on moveout velocities, anisotropy, or a combination of
the two phenomena. In Figure 11.7-5, a close-up display
of the zone of interest between 1.5-2 s shows the overcorrected events with a distinct moveout behavior that
is not quite the same as what a typical overcorrected
event with hyperbolic moveout looks like. Specically,
we see that the events are nearly at with no moveout
within the near-oset range, while the moveout is like
the end of a hockey stick within the far-oset range.
Try attening the events by experimenting with
various velocity picks as shown in Figure 11.-7.6. Note
that, whatever the pick you make on the velocity spectrum, the hockey-stick events do not quite atten;
there is always some curvature left along the moveoutcorrected event.
Equation (11-92) indicates that, for anisotropic velocity analysis, we need to scan two parameters, vN M O
and . Note that the parameter is only present in
the fourth-order moveout term that is signicant at far
osets. This suggests a two-stage parameter scan as described below.

FIG. 11.7-1. (a) Application of Huygens principle to


anisotropic plane-wave propagation from an exploding reector (after Uren et al., 1990a). (b) The isotropic and
anisotropic wavefront associated with an expanding P -wave
(after Sheri, 1991.) See text for details.

(d) The NMO velocity measured from the slope of the


t2 x2 curve indicates that anisotropy makes the
velocity oset dependent (Figure 11.7-4b).
Figure 11.7-5 shows a CMP gather and its velocity spectrum computed using the isotropic traveltime
equation (11-90). Note that there are some overcor-

(a) Perform hyperbolic velocity analysis using only the


rst two terms on the right-hand side of equation (11-92) and the near osets where there is no
hockey-stick eect. Figure 11.7-7 shows the same
CMP gather as in Figure 11.7-5, but with far-oset
traces muted, and the associated velocity spectrum. This rst-stage analysis would give an estimate of the moveout velocity vN M O as in equation
(11-92).
(b) Next, insert the estimated vN M O function into
equation (11-92) and compute the spectrum as
shown in Figure 11.7-8. After picking an function
in time, apply the fourth-order moveout correction
given by equation 11-92 to the CMP gather.
Figures 11.7-9 and 11.7-10 show portions of a
CMP stacked section with and without accounting for
anisotropy in velocity analysis and moveout correction.
The portion shown in Figure 11.7-9 is abundantly rich
in diractions and the portion shown in Figure 11.7-10
contains steeply dipping fault-plane reections conicting with gently dipping reections. Panel (a) in both
gures shows the full-fold stack and panel (b) shows
the near-oset stack based on isotropic velocity analysis and moveout correction using the rst two terms of
equation (11-92). Note that the full-fold stack has been

1968

Seismic Data Analysis


vN M O () for a dipping reector to the moveout velocity vN M O (0) = 0 for a at reector.
While the DMO process corrects for the dip effect on moveout velocities as described by equation
(11-93), it also maps common-oset data, which have
been moveout-corrected using the at-event velocities
vN M O (0), to zero oset. Recall from Section 5.1 that
this mapping is achieved by an integral transform given
by equation (5-14a) where the factor A of equation (5-5)
in the current notation is given by
A=

FIG. 11.7-2. Group velocity functions for P - and SHwaves in Green River shale (After Uren et al., 1990b; Thomsen, 1986).

adversely aected by the hockey-stick eect on moveout


at far osets, while the near-oset stack has better preserved the diractions and dipping events. Panel (c) of
both Figures 11.7-9 and 11.7-10 shows the full-fold stack
derived from anisotropic velocity analysis and moveout
correction using both the second-order and fourth-order
terms in equation (11-92). Note that the full-fold stacking based on anisotropic moveout correction has better
preserved the diractions and dipping events compared
to the full-fold stacking based on isotropic moveout correction. These observations are more evident on the migrated sections shown in Figures 11.7-11 and 11.7-12.

Anisotropic Dip-Moveout Correction


Since anisotropy directly inuences propagation velocity, it certainly should have an impact on dipmoveout correction. Recall from Section 5.1 that 2-D
dip-moveout (DMO) correction removes the dip eect
on moveout velocities, and from Section 7.2 that 3D dip-moveout correction removes both the dip and
source-receiver azimuth eects on moveout velocities.
Rewrite the normalized moveout velocity relation for
the 2-D case (Levin, 1971) using the current notation
vN M O ()
1
=
,
vN M O (0)
cos

h2 2 sin
t2n vN M O (0)

(11 94)

Here, tn is the event time after NMO correction using


the at-event velocity vN M O , h is the half-oset and
is the reector dip.
As discussed in Section 5.1, the integral transform
factor A given by equation (11-94) requires knowledge
of the reector dip to perform the DMO correction. To
circumvent this requirement, we use the relation (Section D.1)
sin =

vN M O (0)ky
,
20

(11 95)

which states that the reector dip can be expressed


in terms of midpoint wavenumber ky and frequency
0 the Fourier duals of midpoint y0 and zero-oset
event time 0 associated with the DMO-corrected data,
respectively. Substitute equation (11-95) into equation
(11-94) to get
A=

1+

h2 ky2
.
t2n 02

(11 96)

The frequency-wavenumber domain dip-moveout


correction that transforms the normal-moveout-corrected prestack data with a specic oset 2h from yn tn
domain to y0 0 domain is achieved by the integral of
equation (5-14a) where A is given by equation (11-96).
We now examine the DMO mapping, specically
the transform factor A of equation (11-94) for the case
of an anisotropic medium. Consider a dipping reector
situated in an anisotropic medium with its symmetry
axis not coincident with the normal to the reector. For
this general case, the normal-moveout velocity is given
by (Tsvankin, 1995)

(11 93)

and note that, from the standpoint of velocities,


isotropic DMO correction maps the moveout velocity

1+

vN M O () =

()
cos

1+

1 d2 ()
() d2

tan d()
1
()
d

(11 97)

Reservoir Geophysics

1969

FIG. 11.7-3. Crossplot of P -wave anisotropy parameters, and , associated with various rock samples (after Thomsen,
1986).

1970

Seismic Data Analysis


elliptical ( = 0), note that equation (11-98) reduces
to the isotropic case given by equation (11-93).
Compare equations (11-93) and (11-98) and note
that, for the case of > 0, the isotropic DMO correction factor is smaller than the anisotropic DMO correction factor. This means that an anisotropic DMO correction would have a larger aperture than the isotropic
DMO correction. However, depending on the magnitude
of the anisotropy parameters, and , and their signs,
the opposite could also be true. In fact, setting = in
equation (11-98), the case of elliptical anisotropy, the
eect of anisotropy on dip-moveout cancels out altogether.
We are now ready to redene the transform factor
A of equation (11-94) for anisotropic DMO correction
(Anderson and Tsvankin, 1997). To use the anisotropic
moveout velocity of equation (11-98), rst, rewrite equation (11-94) as
A=

1+

4h2 1 cos2
.
2
t2n vN
M O (0)

(11 100a)

Then substitute equation (11-93) to get the desired expression


A=
FIG. 11.7-4. (a) CMP raypaths associated with a at reector in a transversely isotropic medium; (b) the t2 x2
plot that shows the oset-dependent behavior of the NMO
velocity (After Uren et al., 1990c).

where is the phase angle measured from the vertical,


is the reector dip (Figure 11.7-2), and is the phase
velocity evaluated in the direction coincident with the
reector dip, = .
By using equation (11-84) for the phase velocity,
Tsvankin (1995) computes the derivatives in equation
(11-97) and derives an expression for the moveout velocity associated with the case of weak anisotropy. Below,
we write his equation in normalized and compact form
vN M O ()
1
=
1 + ( ) B ,
vN M O (0)
cos

(11 98)

where
B=

sin3
4sin4 9 sin2 + 6 .
1 sin2

(11 99)

Equation (11-98) states that anisotropic DMO correction maps the moveout velocity vN M O () for a dipping reector to the moveout velocity vN M O (0) for a
at reector. By setting the anisotropy parameters =
= 0, or making the assumption that the anisotropy is

1+

1
4h2
1
. (11 100b)
2
2
2
tn vN M O (0) vN M O ()

By substituting equations (11-98) and (11-95) into


equation (11-100b), and making the weak anisotropy
assumption, we obtain the transform factor given by
Anderson and Tsvankin (1997), which we write below
in compact form,
A=

1+

h2 ky2
1+ C ,
t2n 02

(11 101)

where
C = 8 sin4 18 sin2 + 12,

(11 102)

and sin is given by equation (11-95) in terms of


vN M O (0), ky , and 0 .
To implement the frequency-wavenumber domain
anisotropic dip-moveout correction, again, use the integral of equation (5-14a) where A is given by equation
(11-101).
By setting the anisotropy parameters = = 0,
or making the assumption that the anisotropy is elliptical ( = 0), note that equation (11-101) reduces
to the isotropic case given by equation (11-96). The implementation of anisotropic DMO correction requires
a simple modication to the isotropic implementation,
given by the factor [1 + ( )C]. Note that both for
the moveout velocity given by equation (11-98) and the
DMO transform factor given by equation (11-101), it is
the dierence , and not the individual anisotropy

Reservoir Geophysics
parameters, that dictates the anisotropic eect. What
remains to be determined is the dierence . Combine equations (11-87) and (11-91) to get the relation
=

vN M O (0)
0

(11 103)

The eective anisotropy parameter and the


at-reector moveout velocity vN M O (0) are estimated
by anisotropic velocity analysis described earlier in
this section (equation 11-92). The scaling velocity 0
is associated with a vertically incident P -wave in
an isotropic, horizontally layered earth. Finally, note
that isotropic DMO correction (equation 11-96) is
velocity-independent, whereas anisotropic DMO correction (equation 11-101) is velocity-dependent by way of
equation (11-102).
The anisotropic DMO impulse response can vary
in shape and depart from the familiar elliptical shape
associated with the isotropic DMO impulse response.
Figure 11.7-13 shows some of the anisotropic impulse
responses created by Anderson and Tsvankin (1997) using a frequency-wavenumber domain DMO correction
based on the above formulation. For = 0, the
case of elliptical anisotropy, the impulse response kinematically is identical to the isotropic impulse response.
For > 0, the aperture of the anisotropic impulse
response broadens while maintaining the upward curvature, and for < 0, the curvature is reversed
downward.
An interesting theoretical conjecture can be drawn
from the behavior of isotropic and anisotropic DMO
impulse responses shown in Figures 11.7-13a,b. While
anisotropy causes the DMO aperture to be broadened,
vertically increasing velocity causes the DMO aperture
to be narrowed (Section 5.1). As a result of these two
counteracting eects, we may conclude that anisotropic
DMO coorection that accounts for vertically varying velocity may be equivalent to constant-velocity isotropic
DMO correction. Does this mean that, for most eld
data cases where velocities vary vertically, it suces
to perform constant-velocity isotropic DMO corection
even in the presence of anisotropy? This would implicitly account for any anisotropic behavior that might be
present in the data. In fact, Levin (1990) conducted numerical studies of reection times from a dipping plane
in a transversely isotropic media using elastic parameters associated with four dierent sandstone, shale, and
limestone samples, and concluded that, when the symmetry axis of transverse anisotropy is perpendicular to
the reector, the isotropic moveout velocity given by
equation (11-93) is largely valid. Larner (1993) extended
this work to the case of a medium with vertically varying velocity and reached a similar conclusion that, for

1971

most cases of weak anisotropy, isotropic DMO correction is largely valid.


We refer to the eld data example shown in Figure 11.7-14. Following isotropic DMO correction, the
CMP gather as in Figure 11.7-5 exhibits the hockeystick behavior of the anisotropic moveout more distinctly. Whatever the velocity pick, the isotropic moveout correction using equation (11-90) fails to atten the
events within the time gate 1.5-2 s (Figure 11.7-15).
By rst applying isotropic moveout correction using the
near-oset traces in the gather, we derive an estimate of
the normal-moveout velocity vN M O (0) as shown in Figure 11.7-16. This then is followed by the analysis of the
anisotropy parameter and applying the fourth-order
moveout correction using equation (11-92) (Figure 11.717). Note that, despite the DMO correction that did
not account for anisotropy, the subsequent anisotropic
moveout correction has been successful in attening the
events almost completely.
Scan the anisotropic velocity analyses shown in
Figures 11.7-18 through 11.7-23 to examine the extent
of anisotropy manifested by these DMO-corrected gathers and how well the anisotropy has been accounted for
by the post-DMO velocity analysis. The failure in attening the events completely in some gathers may stem
from various sources of error associated with the assumptions made about anisotropy, and it may even be
caused by not correcting for anisotropy during DMO
correction that preceded the velocity analysis.
Figures 11.7-24 and 11.7-25 show portions of a
stacked section with isotropic DMO correction which
was followed by velocity analysis with and without
anisotropy accounted for. Compare with the corresponding panels in Figures 11.7-9 and 11.7-10 and note
that the steep fault-plane reections and diraction
anks have been preserved by DMO correction. In fact,
much of the task of preserving steep dips with conicting dips already has been achieved by isotropic
dip-moveout processing as seen in panel (a) in both
Figures 11.7-24 and 11.7-25 that shows the full-fold
stack based on isotropic velocity analysis and moveout
correction using the rst two terms of equation (1192). For comparison, panel (b) in both Figures 11.724 and 11.7-25 shows the near-oset stack based on
isotropic processing. Panel (c) of both Figures 11.724 and 11.7-25 shows the full-fold DMO stack derived from post-DMO anisotropic velocity analysis and
moveout correction using both the second-order and
fourth-order terms in equation (11-92). Compared with
panel (a) in the same gures, the dierence between
the isotropic and anisotropic processing may be viewed
at best as marginal. Again, this inconclusive result
is attributable to the fact that DMO correction was
done without accounting for anisotropy. The marginal
(text continues on p. 1976)

1972

Seismic Data Analysis

FIG. 11.7-5. (a) An NMO-corrected CMP gather and (b) its velocity spectrum based on isotropic moveout equation (11-90);
(c) and (d) are enlargements of (a) and (b), respectively.

Reservoir Geophysics

1973

FIG. 11.7-6. (a) The CMP gather and (b) its velocity spectrum as in Figure (11.7-5c,d), except that the NMO correction
has been applied using a dierent velocity function.

1974

Seismic Data Analysis

FIG. 11.7-7. (a) The NMO-corrected CMP gather as in Figure 11.7-5a but with a dierent mute, and (b) its velocity
spectrum; (c) and (d) are enlargements of (a) and (b), respectively.

Reservoir Geophysics

1975

FIG. 11.7-8. (a) The NMO-corrected CMP gather as in Figure 11.7-5c and (b) its -spectrum computed from the anisotropic
moveout equation (11-92); (c) the same gather as in (a) after anisotropic moveout correction using the function picked from
the spectrum in (d).

1976

Seismic Data Analysis

Reservoir Geophysics

1977

1978

Seismic Data Analysis

Reservoir Geophysics

1979

1980

Seismic Data Analysis


In practice, the moveout associated with an event observed in a CMP gather actually is a combination of all
the contributions of the moveout eects listed above. It
makes sense in practice to attempt to break up the total moveout into individual components and correct for
them one at a time. A pragmatic workow for moveout
correction is NMO correction (moveout type (a)), followed by DMO correction (combined moveout types (b)
and (c)), and anisotropic moveout correction (moveout
type (d)). The 2-D eld data example shown in Figures
11.7-26 and 11.7-27 is based on this moveout correction
sequence with the exclusion of moveout type (c) related
to 3-D data.

Anisotropic Migration
A convenient way to understand the eect of anisotropy
on migration is to refer to the zero-oset dispersion relation of the one-way scalar wave equation that describes
wave propagation associated with exploding reectors
(Section 4.1), using the present notation,
kz =

FIG. 11.7-13. DMO impulse responses for three dierent


cases of the parameter: (a) = 0, the case of elliptical
anisotropy which is kinematically identical to the isotropic
DMO impulse response, (b) > 0, and (c) < 0. (after
Anderson and Tsvankin, 1997.)

dierences are also evident on the migrated sections


shown in Figures 11.7-26 and 11.7-27. By accounting
for anisotropy in DMO correction, however, the subsequent imaging of the steeply dipping fault-plane reections can be improved (Alkhalifah et al., 1996).
Throughout the various stages in data analysis, we
encounter dierent moveout types:
(a) Normal moveout associated with at events,
(b) Dip moveout associated with dipping events,
(c) Azimuthal moveout associated with 3-D recording
geometry of varying source-receiver directions, and
(d) Anisotropic moveout caused by directional changes
in velocity.

2
0

0 ky
2

(11 104)

where kz and ky are the wavenumbers associated with


the depth z and midpoint y axes, is the temporal
frequency and is the medium velocity used to migrate a zero-oset compressional waveeld. The dispersion relation given by equation (11-104) is the basis from
which all zero-oset nite-dierence and frequencywavenumber migration algorithms are developed (Sections D.1, D.4, and D.7). Hence, it makes good sense to
examine what form this dispersion relation takes in the
case of an anisotropic medium.
Accompanying the dispersion relation is the equation for wave extrapolation used in nite-dierence and
frequency-wavenumber algorithms (Section 4.1). Given
a zero-oset upcoming waveeld P (y, z = 0, t) at the
surface z = 0, for which a stacked section is a good
approximation, the objective is to extrapolate it downward at depth steps of z using the extrapolation equation
P (ky , z + z, ) = P (ky , z, ) exp (ikz z).
(11 105)
This is then followed by invoking the imaging principle
which states that the migrated section P (x, z, t = 0) at
each depth z corresponds to the summation over the
frequency components of the extrapolated waveeld, or
equivalently collecting the extrpolated waveeld at time
t = 0.
(text continues on p. 1995)

Reservoir Geophysics

1981

FIG. 11.7-14. (a) The NMO-corrected CMP gather as in Figure 11.7-5a after DMO correction and (b) its velocity spectrum
based on isotropic moveout equation (11-90); (c) and (d) are enlargements of (a) and (b), respectively.

1982

Seismic Data Analysis

FIG. 11.7-15. (a) The DMO-corrected CMP gather and (b) its velocity spectrum as in Figure (11.7-14c,d), except that the
NMO correction has been applied using a dierent velocity function.

Reservoir Geophysics

1983

FIG. 11.7-16. (a) The NMO-corrected CMP gather as in Figure 11.7-14a after DMO correction but with a dierent mute,
and (b) its velocity spectrum; (c) and (d) are enlargements of (a) and (b), respectively.

1984

Seismic Data Analysis

FIG. 11.7-17. (a) The NMO-corrected CMP gather as in Figure 11.7-14c after DMO correction and (b) its -spectrum
computed from the anisotropic moveout equation (11-92); (c) the same gather as in (a) after anisotropic moveout correction
using the function picked from the spectrum in (d).

Reservoir Geophysics

1985

FIG. 11.7-18. (a) A portion of an NMO-corrected CMP gather after DMO correction and (b) its -spectrum computed
from the anisotropic moveout equation (11-92); (c) the same gather as in (a) after anisotropic moveout correction using the
function picked from the spectrum in (d).

1986

Seismic Data Analysis

FIG. 11.7-19. (a) A portion of an NMO-corrected CMP gather after DMO correction and (b) its -spectrum computed
from the anisotropic moveout equation (11-92); (c) the same gather as in (a) after anisotropic moveout correction using the
function picked from the spectrum in (d).

Reservoir Geophysics

1987

FIG. 11.7-20. (a) A portion of an NMO-corrected CMP gather after DMO correction and (b) its -spectrum computed
from the anisotropic moveout equation (11-92); (c) the same gather as in (a) after anisotropic moveout correction using the
function picked from the spectrum in (d).

1988

Seismic Data Analysis

FIG. 11.7-21. (a) A portion of an NMO-corrected CMP gather after DMO correction and (b) its -spectrum computed
from the anisotropic moveout equation (11-92); (c) the same gather as in (a) after anisotropic moveout correction using the
function picked from the spectrum in (d).

Reservoir Geophysics

1989

FIG. 11.7-22. (a) A portion of an NMO-corrected CMP gather after DMO correction and (b) its -spectrum computed
from the anisotropic moveout equation (11-92); (c) the same gather as in (a) after anisotropic moveout correction using the
function picked from the spectrum in (d).

1990

Seismic Data Analysis

FIG. 11.7-23. (a) A portion of an NMO-corrected CMP gather after DMO correction and (b) its -spectrum computed
from the anisotropic moveout equation (11-92); (c) the same gather as in (a) after anisotropic moveout correction using the
function picked from the spectrum in (d).

Reservoir Geophysics

1991

1992

Seismic Data Analysis

Reservoir Geophysics

1993

1994

Seismic Data Analysis

Reservoir Geophysics

1995

for substitution into equation (11-106b) to get


ky2 kz2 + ky2
4 2
2
1
+
=

0
2
ky2 + kz2
ky2 + kz2

FIG. 11.7-28. Wavefronts for an expanding P -wave in an


anisotropic medium for two dierent combinations of Thomsen parameters and : (a) = 0 and (b) > 0 (after
Thomsen, 1986).

To account for the eect of anisotropy in migration, the medium velocity 0 in the dispersion relation
given by equation (11-104) needs to be replaced with the
P -wave phase velocity given by equation (11-84). This
means that velocity used in anisotropic migration is dependent on the phase angle of the propagating wavefront (Figure 11.7-1). We shall rst rewrite the dispersion relation of equation (11-104) explicitly in terms of
the velocity
4 2
= 02 ,
+ kz2

(11 106)

ky2

and replace 0 with the phase velocity of equation (1184) specied for the reector dip
4 2
2
= 02 1 + sin2 cos2 + sin4 .
ky2 + kz2
(11 107)
Next, we shall use the relations (Section D.1)
sin =

ky
ky2

(11 108a)

+ kz2

and
cos =

kz
ky2 + kz2

(11 108b)

(11 109)

By setting the anisotropy parameters = = 0, note


that the anisotropic dispersion relation given by equation (11-109) reduces to the isotropic case given by
equation (11-106).
A straightforward practical implementation of the
anisotropic dispersion relation of equation (11-109) is
within the framework of a frequency-wavenumber algorithm such as Stolt or phase-shift migration. By using
an anisotropic dispersion relation dened for time migration, Anderson et al. (1996) have modied Fowler
DMO and prestack time migration (Section 5.3) to
account for anisotropy. Nevertheless, various practical forms of the dispersion relation can also be used
to include the eect of anisotropy in frequency-space
nite-dierence explicit or implicit schemes (Kitchenside, 1993; Ristow and Ruehl, 1997). Although it is
only a theoretically interesting conjecture, elliptical
anisotropy may be accounted for by a vertical stretching in depth that follows an isotropic time migration
(Verwest, 1989).
The eect of anisotropy on an expanding wavefront
is seen in Figure 11.7-28. Depending on the values of the
Thomsen parameters, the wavefront can take various
shapes, sometimes quite warped. These wavefronts can
be considered as the kinematic aspect of an anisotropic
migration impulse response. Note that, at certain dips,
more migration takes place if anisotropy is accounted
for as compared to isotropic migration. On the other
hand, at some other dips, less migration may take place,
depending on the anisotropy parameters.
A widely recognized case of anisotropy is from oshore West Africa. Figure 11.7-29 shows a popular example of anisotropic processing (Alkhalifah et al., 1996).
The data were processed by applying both isotropic and
anisotropic DMO correction and migration. Note the
better imaging of the steep fault planes by accounting
for anisotropy in processing.
Note from the anisotropic dispersion relation given
by equation (11-109) that it is important to supply
an anisotropic migration algorithm with accurate estimates of the anisotropy parameters and . It is most
likely safer to do isotropic migration than anisotropic
migration with incorrect values for and . Alkhalifah
and Larner (1994) conducted some numerical studies
for migration error in transversely isotropic media and
concluded that, fortuitously, isotropic migration in the
presence of anisotropy yields fairly accurate results for
dips up to 50 degrees. For steeper dips, the isotropic

1996

Seismic Data Analysis

FIG. 11.7-29. A portion of a 2-D seismic section based on the application of DMO correction and phase-shift time migration
(a) without and (b) with anisotropy accounted for (Alkhalifah et al., 1996).

Reservoir Geophysics

1997

FIG. 11.7-30. P -to-P reection amplitudes as a function of angle of incidence at two dierent interfaces associated with a
horizontally layered earth model with transverse isotropy (Haase, 1998). See text for details.

1998

Seismic Data Analysis

FIG. 11.7-31. Shear-wave splitting in an azimuthally anisotropic medium.

assumption no longer is valid. Nevertheless, Alkhalifah


and Larner (1994) also caution the practitioner of the
undesirable consequences of anisotropic migration using
poorly estimated and parameters.

Eect of Anisotropy on AVO


The P -to-P reection amplitude as a function of angle
of incidence given by equation (11-16) can be modied to accommodate for transverse isotropy as follows
(Tsvankin, 1996; Rueger, 1997)
R() =

1
2

2
2 1
1
4 2
2 2
+ sin2
2


2

1 1
+
2
2

tan2 sin2 ,

(11 110)

where and are changes in anisotropy across


the at interface that separates the upper and lower
anisotropic media. By setting = = 0, equation
(11-110) reduces to the isotropic form given by equation (11-16). Since the origin of equation (11-16) is the
Aki-Richards equation (11-15), equation (11-110) also is
based on the assumption that, in addition to the small
changes in elastic parameters, the changes in anisotropy
parameters and are also small across the interface.

Reservoir Geophysics

1999

FIG. 11.7-32. Shear-wave splitting from a North Sea 3-D survey: (a) radial and (b) transverse components as a function of source receiver-azimuth in degrees. See text for details. (Figure courtesy Gaiser, 1999a, and Baker-Hughes Western
Geophysical.)

The terms of equation (11-110) can be split into


two parts, R() = Ri () + Ra () the isotropic component Ri () identical to the terms of equation (11-16),
and the anisotropic component Ra () given by (Rueger,
1997; Haase, 1998)
1
1
sin2 + tan2 sin2 . (11 111)
2
2
Figure 11.7-30 shows P -to-P reection amplitudes as
a function of angle of incidence at two dierent interfaces associated with a horizontally layered earth model
with transverse isotropy (Haase, 1998). The curves labeled as Ri correspond to the isotropic component
given by equation (11-16) for three dierent values of
/ ratios and the curves labeled as Ra correspond to
the anisotropic component of the reection amplitudes
given by equation (11-111) for a specic combination of
and values. Since Ra () does not depend on /,
the Ra () curve is common to each of the three cases
of the Ri () curves. To get the total response given by
equation (11-110), add the curves associated with the
two components.
Ra () =

There are two important eects of anisotropy on


AVO:
(a) The polarity of P -to-P reection amplitudes can
reverse for some combination of the anisotropy parameters.
(b) The anisotropic eect on reection amplitudes is
most signicant at large angles of incidence.

Shear-Wave Splitting in Anisotropic Media


In Section 11.6, we reviewed the 4-C seismic method
based on the conversion of P -waves to S-waves at layer
boundaries. The converted S-wave propagating in an
anisotropic medium exhibits a unique behavior. Consider a rock layer that contains vertical fractures oriented in, say the north-south direction, as sketched in
Figure 11.7-31 a case of a horizontally transverse
isotropy, or equivalently, a case of azimuthal anisotropy.
The S-wave that enters this transversely isotropic layer
from below emerges from above as split and polarized

2000

Seismic Data Analysis


The split shear waves interfere with one another
in a way that depends on the source-receiver azimuth.
Figure 11.7-32 shows radial and transverse components
sorted with respect to source-receiver azimuth. Each
trace in these gathers is associated with a radial or
transverse component in a specic azimuthal direction.
Note that the radial component is fairly consistent from
one azimuthal direction to the next. The transverse
component, on the other hand, shows polarity change
every 90 degrees. From these gathers, the Alford rotation produces the fast and slow shear-wave components
while minimizing the energy in the transverse component.

EXERCISES

FIG. 11.E-1. The moveout-corrected CMP gather referred


to in Exercise 11-6.

in two orthogonal directions the shear-wave component that is polarized parallel to the fracture orientation
and the shear-wave component that is polarized perpendicular to the fracture orientation. Shear-wave splitting
in azimuthally anisotropic media is sometimes referred
to as shear-wave birefringence.
The shear-wave component polarized parallel to
the fracture orientation travels with a speed faster than
the shear-wave component in the perpendicular direction. In practice, we have the radial and transverse components; but not the slow and fast shear-wave components. The fast and slow shear-wave components, however, can be extracted from the radial and transverse
geophone components (Section 11.6) by a special rotation developed by Alford (1986). The Alford rotational
analysis of the shear-wave data provides an estimate
of fracture orientation. This is strategically important
in developing reservoirs within fractured rocks (Martin
and Davis, 1987; Ata and Michelena, 1995); specically,
in planning trajectories for production wells.

Exercise 11-1. Derive equation (11-2a) for the


Fresnel zone using the geometry of Figure 11.1-3.
Exercise 11-2. Refer to Figure 11.4-1a. Consider a
surface multiple from the rst reecting interface. Trace
the traveltime on the VSP diagram in Figure 11.4-1b.
Multiples do not reach the downgoing wave path; thus,
they can be eliminated by corridor stacking.
Exercise 11-3. Refer to Figure 11.4-1b. Should
the slopes of the downgoing and upcoming waves associated with a layer be the same in magnitude?
Exercise 11-4. What procedure does CMP stacking correspond to in the f k domain?
Exercise 11-5. Sketch the traveltime response for
a point scatterer on a zero-oset VSP record.
Exercise 11-6. Consider a CMP gather with a
single reection event. Suppose you have applied hyperbolic moveout correction using equation (11-90) and
discovered that the event is not at for all osets. Instead, the moveout-corrected event may have one of the
three shapes shown in Figure 11.E-1. Match each of the
curves A, B, and C with the following three possibilities:
(a) You have applied hyperbolic moveout correction
using an erroneously low velocity in equation (1190).
(b) You have applied a second-order moveout correction (equation 3-4b) to an event that has a fourthorder moveout behavior described by equation (35a).
(c) You have ignored anisotropy in moveout correction
described by equation (11-92).

Appendix L
MATHEMATICAL FOUNDATION OF ELASTIC WAVE PROPAGATION

L.1 Stress-Strain Relation


Seismic waves induce elastic deformation along the propagation path in the subsurface. The
equation of wave propagation in elastic solids are derived by using Hookes law and Newtons
second law of motion. We shall begin with the stress-strain relation for elastic solids.
Solid bodies, such as rocks are capable of propagating forces that act upon them. Imagine
an innitesimally small volume around a point within a solid body with dimensions (dx, dy, dz)
as depicted in Figure L-1. Stress is dened as force per unit area. The stress acting upon one
of the surfaces, say dy dz, can in general be at some arbitrary direction. It can, however,
be decomposed into three components: Pxx which is normal to the surface, and Pxy and Pxz
which are tangential to the surface. The rst subscript refers to the direction of the normal to the
surface, and the second subscript refers to the direction of the stress component. The components
of the stress that are normal to the surfaces associated with the innitesimal volume shown in
Figure L-1 are called the normal stress components and the components of the stress that are
tangential to the surfaces are called the shear stress components. A normal stress component is
tensional if it is positive, and compressional if it is negative. To retain the solid volume depicted
in Figure L-1 in equilibrium requires nine stress components that make up the stress tensor

Pxx Pxy Pxz


{Pij } = Pyx Pyy Pyz .
(L 1)
Pzx Pzy Pzz
The diagonal elements are the normal stress components and the o-diagonal elements are the
shear stress components.
As the dimensions of the volume surrounding a point inside a solid body are made innitesimally smaller, the sum of the moments of all surface forces about any axis must vanish.
This requirement causes Pxy = Pyx , Pxz = Pzx , and Pyz = Pzy , making the stress tensor (L-1)
symmetrical.
Fluids cannot support shear stress. In a uid medium, only one independent stress component remains, since Pxx = Pyy = Pzz = p, the hydrostatic pressure (the negative sign signies
the compressive nature of the hydrostatic pressure), and Pxy = Pxz = Pyz = 0.
Stress induces deformation of solid bodies. Consider two points, P and Q, within a solid
body as indicated in Figure L-2. Subject to a stress eld, the solid is deformed in some manner
and the particles at points P and Q are displaced to new locations P and Q . The displacement
components u, v, and w can be expressed by the relation



u
u/x u/y u/z
x
v = v/x v/y v/z y .
(L 2)
w
w/x w/y w/z
z
To understand the physical meaning of the elements of the tensor containing the partial
derivatives in equation (L-2), instead of arbitrary displacement as shown in Figure L-2, consider
deformations of specic types. The simplest deformation is an extension in one direction as a
result of a tensional stress as illustrated in Figure L-3a. (For simplicity, only one surface of the
volume in Figure L-1 is considered.) The fractional change in length in the x-direction is u/x,

2002

Seismic Data Analysis

FIG. L-1. Stress components acting on an innitesimally small volume surrounding a point within an
elastic solid.

FIG. L-2. Two points, P and Q, within a solid body subject to a stress eld. The solid is deformed in
some manner and the particles at points P and Q are displaced to new locations P and Q . Here, u,
v, and w are the displacement components.

which, in the limit as the volume becomes innitesimally small, is dened as the normal strain
component.
Strain is a dimensionless quantity. There are three normal strain components, exx , eyy , and
ezz :
exx =

u
,
x

(L 3a)

eyy =

v
,
y

(L 3b)

w
.
(L 3c)
z
The stress eld away from the typical seismic source is so small that it does not cause any
permanent deformation on rock particles along the propagation path. Hence, the strain induced
ezz =

Mathematical Foundation of Elastic Wave Propagation

2003

FIG. L-3. Deformations caused by stress acting on one surface of the volume in Figure L-1: (a) linear
deformation that results in extension of the side AB in the x direction by an amount BB ; (b) shearing
only; (c) rotation only; (d) combined angular deformation () and rotation (). See text for details.

by seismic waves is very small, usually around 106 . A positive strain refers to an extension and
a negative strain refers to a contraction.
Other types of deformation are caused by shearing (Figure L-3b), rotation (Figure L-3c),
and a combination of the two (Figure L-3d). The angular deformations and , in the limit as
the volume becomes innitesimally small, can be expressed in terms of displacement components
u and w, by using the relations
=

w
x

(L 4a)

+ =

u
,
z

(L 4b)

and

2004

Seismic Data Analysis


which, by adding both sides, yield
= exz =

1 w u
= ezx .
+
2 x
z

(L 5a)

Similarly, we obtain
exy =

1 v
u
= eyx
+
2 x y

(L 5b)

ezy =

1 w v
+
2 y
z

(L 5c)

and

The quantitites described by equations (L-3) and

exx exy
{eij } = eyx eyy
ezx ezy

= eyz .

(L-5) make up the strain tensor

exz
eyz .
(L 6)
ezz

Note that, by way of equations (L-5a,b,c), the strain tensor (L-6) is symmetric.
Also from equations (L-4a,b), we obtain by subtraction
= xz =

1 w u
= zx .

2 x
z

(L 7a)

Similarly,
u
1 v
= yx

2 x y

(L 7b)

1 w v

2 y
z

(L 7c)

xy =
and
yz =

= zy .

Note that by denition


xx = yy = zz = 0.
The angular deformations given by equations (L-5a,b,c) are called shear strains since they
result in a shearing of the volume around a point within a solid body (Figure L-3b). The quantities given by equations (L-7a,b,c) are associated with rotation without deformation (Figure
L-3c).
The displacement tensor in equation (L-2) can be expressed as

u/x u/y u/z


u/x + u/x v/x + u/y w/x + u/z
1
v/x v/y v/z = u/y + v/x v/y + v/y w/y + v/z
2
v/x v/y v/z
u/z + w/x v/z + w/y w/z + w/z

0
v/x u/y w/x u/z
1
u/y v/x
0
w/y v/z .

2
u/z w/x v/z w/y
0
(L 8)
The elements of the matrices in equation (L-8) can then be expressed in terms of the normal
strain components exx , eyy , and ezz , given by equations (L-3a,b,c), the shear strain components,
exz , exy , and eyz , given by equations (L-5a,b,c), and the rigid rotational components xz , xy , and
yz , given by equations (L-7a,b,c). All of these components are then substituted into equation
(L-2) to get



u
exx exy exz
0 xy xz
x
v = eyx eyy eyz + yx
0 yz y .
(L 9)
w
ezx ezy ezz
zx zy
0
z

Mathematical Foundation of Elastic Wave Propagation

2005

Next, we need to establish the relation between the stress tensor (L-1) and the strain
tensor (L-6). For an elastic solid, Hookes law states that the strain at any point is directly
proportional to the stresses applied at that point. First, consider a volume (Figure L-1) under
tensional principle stresses, Pxx , Pyy , and Pzz . The stress-strain relations are written as
Pxx = Eexx ,

(L 10a)

Pxx = Eeyy ,

(L 10b)

Pxx = Eezz ,

(L 10c)

and
where E and are proportionality constants, which are specic to the material and are called
Youngs modulus and Poissons ratio, respectively.
Consider a cylindrical rod that is subjected to a longitudinal extension in the x-direction. All
other stresses being zero, this causes a lateral contraction in the y-direction. Note from equation
(L-10a) that Youngs modulus is the ratio of the longitudinal stress Pxx to the longitudinal
strain exx . Substitute Pxx from equation (L-10a) into equation (L-10b) and note that Poissons
ratio is the ratio of the lateral contraction dened by the strain component eyy to longitudinal
extension dened by the strain component exx . Since strain is a dimensionless quantity, Youngs
modulus has the dimensions of stress, and Poissons ratio is a pure number.
Similar to equations (L-10a,b,c), we establish the following relations
Pyy = Eeyy ,

(L 11a)

Pyy = Eexx ,

(L 11b)

Pyy = Eezz ,

(L 11c)

Pzz = Eezz ,

(L 12a)

Pzz = Eexx ,

(L 12b)

Pzz = Eeyy .

(L 12c)

and

By combining equations (L-10), (L-11), and (L-12), we rewrite the principal stress-strain
relations as
Pxx Pyy Pzz = 3Eexx ,

(L 13a)

Pxx Pyy Pzz = 3Eeyy ,

(L 13b)

Pxx Pyy Pzz = 3Eezz .

(L 13c)

By rewriting equations (L-13a,b,c), we get


(1 + )Pxx (Pxx + Pyy + Pzz ) = Eexx ,

(L 14a)

(1 + )Pyy (Pxx + Pyy + Pzz ) = Eeyy ,

(L 14b)

(1 + )Pzz (Pxx + Pyy + Pzz ) = Eezz ,

(L 14c)

and adding them, we obtain


(1 2)(Pxx + Pyy + Pzz ) = E(exx + eyy + ezz ).

(L 15)

2006

Seismic Data Analysis


Refer to Figure L-1. If the unstrained volume is (xyz), and the strained volume is
(x + u)(y + v)(z + w), then the fractional change in volume, , is
= (x + u)(y + v)(z + w) (x y z) /(x y z).
By neglecting the higher-order terms in this ratio, and referring to the denitions of the principal
strain components given by equations (L-3a,b,c), we nd that the fractional change in volume,
or dilatation , in the limiting case of the volume becoming innitesimally small, is given by
= exx + eyy + ezz ,

(L 16)

which, upon substitution into equation (L-15), yields


Pxx + Pyy + Pzz =

E
.
1 2

(L 17)

Finally, by using the relations (L-16) and (L-17), and back substituting into equations (L13a,b,c), we obtain the relations between the principal stress and principal strain components:
Pxx = + 2exx ,

(L 18a)

Pyy = + 2eyy ,

(L 18b)

Pzz = + 2ezz ,

(L 18c)

where and are the elastic moduli for the solid given by
=

E
(1 + )(1 2)

(L 19a)

E
.
2(1 + )

(L 19b)

and
=

These elastic moduli have the dimensions of stress.


Now consider the volume in Figure L-1 under shear stress components, Pxy , Pxz , and Pyz .
The deformations associated with these stress components are the shear strains, exy , exz , and
eyz . These stress and strain components also are linearly related for elastic solids:
Pxy = 2exy ,

(L 20a)

Pxz = 2exz ,

(L 20b)

Pyz = 2eyz ,

(L 20c)

where the proportionality constant is known as the modulus of rigidity. Note, from equation
(L-20a), that modulus of rigidity is the ratio of shear stress to shear strain.
By combining equations (L-18a,b,c) and (L-20a,b,c), we obtain the stress-strain relation for
elastic solids:

Pxx Pxy Pxz


1 0 0
exx exy exz
Pxy Pyy Pyz = 0 1 0 + 2 exy eyy eyz .
(L 21)
Pzx Pzy Pzz
0 0 1
ezx ezy ezz
This is the formal expression for Hookes law that relates the stress tensor on the left to the
strain tensor on the right. Recall that these two tensors are symmetric, and that the diagonal
elements represent the normal components and the o-diagonal elements represent the shear
components. Equation (L-21) holds for homogeneous, isotropic, elastic solids (for which elastic
behavior is independent of direction), and for deformations that are suciently small (usual
case for seismic waves) to satisfy the linear relationship between stress and strain.

Mathematical Foundation of Elastic Wave Propagation

2007

Since the stress tensor (L-1) and the strain tensor (L-6) are symmetric, equation (L-21) can
be written in the form

Pxx
+ 2

0
0
0
exx
+ 2

0
0
0 eyy
Pyy

+ 2
0
0
0 ezz
Pzz
(L 22)

.
0
0
2
0
0 exy
Pxy 0

Pxz
0
0
0
0
2
0
exz
Pyz
0
0
0
0
0
2
eyz
This equation is a special form of


Pxx
c11
Pyy c21


Pzz c31

=
Pxy c41


Pxz
c51
Pyz
c61

the generalized Hookes law given by (Ocer, 1958)

c12 c13 c14 c15 c16


exx
c22 c23 c24 c25 c26 eyy

c32 c33 c34 c35 c36 ezz


(L 23)

,
c42 c43 c44 c45 c46 exy

c52 c53 c54 c55 c56


exz
c62 c63 c64 c65 c66
eyz

where cij = cji are the 21 independent constants for an elastic medium. The stress vector on
the left is related to the strain vector on the right by the stiness matrix {cij }. Equation (L-23)
states that a stress component is a linear combination of all the strain components. This relation
is the foundation of the linear theory of elasticity and its physical basis is the assumption that
elastic deformations in solids are innitesimally small.
For an isotropic solid, the number of independent constants is reduced to two Lames
constants, and given by equations (L-19a,b); hence equation (L-23) reduces to the special
form of equation (L-22). For a transversely isotropic solid, for which elastic behavior is the same
in two orthogonal directions but dierent in the third direction, the number of independent
constants is ve (Sheri, 1991).

L.2 Elastic Wave Equation


We now review the equations of motion and examine how stress elds are propagated through
an elastic solid. Refer to Figure L-1 and note that, for small displacements and velocities, and
neglecting body forces such as gravity acting on the volume surrounding the points, Newtons
second law of motion gives, for the displacement component u,

Pxx
Pxy
Pxz
2u
=
+
+
,
2
t
x
y
z

(L 24)

where is the density. From the expression (L-21) for Hookes law, we have
Pxx = + 2exx ,

(L 25a)

Pxy = 2exy ,

(L 25b)

Pxz = 2exz .

(L 25c)

and
By making these substitutions into equation (L-24), and assuming that the elastic parameters and are constant, we obtain

exx
exy
exz
2u
=
+ 2
+ 2
+ 2
.
t2
x
x
y
z

(L 26)

By using the denitions for the principal strains given by equations (L-3a,b,c) and that for the
dilatation given by equation (L-16), the latter can be expressed in terms of the displacement

2008

Seismic Data Analysis


components as
=

u v
w
+
+
.
x y
z

(L 27)

Next, by substituting equation (L-27) and the relations for the strain components in terms of
the displacement components given by equations (L-3a,b,c) and (L-5a,b,c) into equation (L-26),
we obtain

2u
w
2u
2u
2u
2w
2v
u v
+
+
+
+
+
2
+

.
=

t2
x x y
z
x2
yx y 2
zx z 2

(L 28a)

Rearrange the terms on the right-hand side to get the following expression

2u
w
u v
2u 2u 2u
+
+
+

=
(
+
)
+ 2 + 2 .
t2
x x y
z
x2
y
z

(L 28b)

Finally, back substitution of equation (L-27) and the denition for the Laplacian operator,
2 : ( 2 /x2 + 2 /y 2 + 2 /z 2 ) into equation (L-28b), yields the equation of motion for the
displacement component u:
2u

+ 2 u.
= ( + )
t2
x

(L 29a)

Similarly, we can derive the equations of motion for the displacement components v and w
2v

= ( + )
+ 2 v
t2
y

(L 29b)

2w

= ( + )
+ 2 w.
2
t
z

(L 29c)

and

get

Dene the displacement vector u :(u, v, w) and combine the three equations (L-29a,b,c) to

2u
= ( + ) + 2 u.
t2

(L 30)

This is the equation of wave propagation in homogeneous, isotropic, and elastic solids.

L.3 Seismic Wave Types Body Waves and Surface Waves


Equation (L-30) can be specialized to describe various wave types that travel within solids and
uids (body waves), and along free surfaces and layer boundaries (surface waves). We shall
derive the equations for two types of body waves that are of interest in exploration seismology
the compressional wave for which the displacement is in the direction of propagation and
the shear wave for which the displacement is in the direction perpendicular to the direction of
propagation. Surface waves are more of an interest in earthquake seismology; as such, we shall
not deal with them here.
To begin with, dierentiate equation (L-29a) with respect to x

2
2u 2u 2u
2u
=
(
+
)
+

+ 2 + 2 ,
x t2
x2
x x2
y
z

(L 31a)

equation (L-29b) with respect to y

2v
y t2

= ( + )

2
2v
2v
2v
+
+ 2+ 2 ,
2
2
y
y x
y
z

(L 31b)

Mathematical Foundation of Elastic Wave Propagation

2009

and equation (L-29c) with respect to z

2w
z t2

= ( + )

2
2w 2w 2w
.
+

+
+
z 2
z x2
y 2
z 2

(L 31c)

Next, reorder the dierentiation on the left-hand side of equations (L-31a,b,c) and add them to
get

2 u v
2 2 2
w
=
(
+
)
+
+
+
+
t2 x y
z
x2
y 2
z 2
2
2
2
2
u u u
v 2v 2v
2w 2w 2w
+
+
+
.
+
+
+
+
+
+
x x2 y 2 z 2
y x2 y 2 z 2
z x2 y 2 z 2
(L 32)

Then, change the order of dierentiation and rearrange the terms on the right-hand side to get

2 u v
2 2
w
=
(
+
)
+
+
+
+
t2 x y
z
x2
y 2
2
2 u v
w
+ 2
+ 2
+
+
x x y
z
y

2
z 2
2 u v
w
w
u v
+ 2
.
+
+
+
+
x y
z
z x y
z
(L 33)

Finally, substitution of equation (L-27) and the denition for the Laplacian operator, 2 :
( 2 /x2 + 2 /y 2 + 2 /z 2 ) into the above expression, yield the equation for dilatational or
compressional-wave propogation:

2
= ( + 2)2 .
t2

(L 34)

The velocity with which compressional waves propagate in an elastic solid then is
=

+ 2
.

(L 35)

Equation (L-34) also can be derived by using the more compact vector algebra. Specically,
by taking the divergence of both sides of equation (L-30) and reordering the dierentiation, we
have

2
( u) = ( + ) () + 2 ( u).
t2

(L 36)

By referring to equation (L-27), note that the dilatation is the divergence of the displacement
vector: = u. Substitution of this relation into equation (L-36) and observing that the
divergence of a gradient is the Laplacian, () = 2 (Lass, 1950), yield equation (L-34)
that describes the compressional-wave propagation.
Dierentiate equation (L-29b) with respect to z

2v
z t2

= ( + )

+
2 v ,
zy
z

(L 37a)

+
2 w .
yz
y

(L 37b)

and equation (L-29c) with respect to y

2w
y t2

= ( + )

Next, change the order of dierentiation and subtract equation (L-37a) from (L-37b)

2 w v

t2 y
z

= 2

w v
.

y
z

(L 38)

2010

Seismic Data Analysis


Finally, substitute equation (L-7c) to get

2 yz
= 2 yz .
t2

(L 39)

Similarly, dierentiate equation (L-29a) with respect to z

2
2u

= ( + )
+
2 u ,
2
z t
zx
z

(L 40a)

and equation (L-29c) with respect to x

2w
x t2

= ( + )

+
2 w .
xz
x

(L 40b)

Next, change the order of dierentiation and subtract equation (L-40a) from (L-40b)

2 w u
w u
= 2
,

t2 x
z
x
z

(L 41)

and substitute equation (L-7a) to get

2 xz
= 2 xz .
t2

(L 42)

Finally, dierentiate equation (L-29a) with respect to y

2
2u

=
(
+
)
+
2 u ,
2
y t
yx
y

(L 43a)

and equation (L-29b) with respect to x

2v
x t2

= ( + )

+
2 v .
xy
x

(L 43b)

Next, change the order of dierentiation and subtract equation (L-43a) from (L-43b)

2 v
v
u
u
= 2
,

t2 x y
x y

(L 44)

and substitute equation (L-7b) to get

2 xy
= 2 xy .
t2

(L 45)

Dene the rotational vector : (xy , xz , yz ), and combine the three equations (L-39), (L-42),
and (L-45) to get the equation for rotational or shear-wave propagation:
2
= 2 .
t2
The velocity with which shear waves propagate in an elastic solid then is

(L 46)

(L 47)

Equation (L-46) also can be derived by using the more compact vector algebra. Specically,
by taking the curl of both sides of equation (L-30) and reordering dierentiation, we have

2
( u) = ( + ) + 2 ( u).
t2

(L 48)

By referring to equations (L-7a,b,c) note that the rotational vector : (xy , xz , yz ) is the curl
of the displacement vector: = (1/2) u. Substitution of this relation into equation (L-48)
and observing that the curl of a gradient is zero, = 0 (Lass, 1950), yield equation (L-46)
for rotational or shear-wave propagation.

Mathematical Foundation of Elastic Wave Propagation

2011

By taking the ratio of equations (L-35) and (L-47), and using the relations (L-19a,b),
we obtain the expression for the ratio of the shear-wave velocity to the compressional-wave
velocity

1 2
,
2(1 )

(L 49)

which indicates that the velocity ratio /, depends on Poissons ratio , and that the
compressional-wave velocity is always greater than the shear-wave velocity .
The wave type described by equation (L-34) is commonly referred to as the P -wave (or
equivalently, longitudinal wave, dilatational wave, or compressional wave), and the wave type
described by equation (L-46) is commonly referred to as the S-wave (or equivalently, rotational
wave, transverse wave, or shear wave). For P -waves, the particle motion is in the direction of
wave propagation, whereas for S-waves, the particle motion is in the direction perpendicular to
the direction of wave propagation.
To understand the nature of P - and S-wave propagation, consider the simple case of planewave propagation in the x-direction. This makes the displacement vector u : (u, v, w) a function
of x only. Refer to equations (L-29a,b,c), substitute for the cubical dilatation from equation
(L-27), and nally, retain only the derivatives with respect to x to obtain

2u
2u
=
(
+
2)
,
t2
x2

(L 50a)

2v
2v
=

,
t2
x2

(L 50b)

and

2w
2w
= 2.
(L 50c)
2
t
x
From equation (L-50a), note that the particle displacement u (the longitudinal vibration)
which is in the direction of propagation (in this case in the x-direction) travels with the velocity
of compressional waves (equation L-35). From equations (L-50b,c), the particle displacements v
and w (the transverse vibrations), which are in the directions perpendicular to the direction of
propagation, travel with the velocity of shear waves (equation L-47).
To a great extent, seismic waves traveling in the earth can be considered as elastic waves.
Therefore, we can use the wave equation (L-30) for elastic solids to describe seismic wave propagation. Source types used in seismic prospecting often generate P -waves. Furthermore, in marine
exploration, because of the water layer surrounding the source, S-waves cannot propagate. Nevertheless, at non-normal incidence, part of the P -wave energy is converted to S-waves. The
P -to-S conversion phenomenon is the basis for the 4-C seismic method (Section 11.6).
Equations for elastic wave propagation can be specialized to describe acoustic wave propagation induced by hydrostatic stress. Hookes law given by equation (L-21) takes the form

Pxx = + 2exx ,

(L 51a)

Pyy = + 2eyy ,

(L 51b)

Pzz = + 2ezz .

(L 51c)

and
Also, note that the normal stress components are equal to the hydrostatic pressure, Pxx =
Pyy = Pzz = P , and that shear stress components vanish, Pxy = Pxz = Pyz = 0. By summing
the three equations (L-51a,b,c) and substituting for the dilatation from equation (L-16), we
obtain
2
(L 52a)
= + ,
3

2012

Seismic Data Analysis


where = P/ is the ratio of the hydrostatic stress P to the volumetric strain and is called
the incompressibility or bulk modulus.
Solve equation (L-52a) for and substitute into equation (L-35) to get the expression for
the compressional-wave velocity in terms of bulk modulus and modulus of rigidity
=

4
+
3 .

(L 52b)

For uids, modulus of rigidity is zero. Hence, by setting = 0, = and using the relation
(L-52), for acoustic waves traveling in uid media, equation (L-34) takes the form

2P
= 2 P,
t2

(L 53)

where P represents the pressure waveeld, which travels with a speed /.


Since the P - and S-waves travel within an elastic solid, they are called body waves. When
an elastic solid is bounded, it can also support waves traveling along the boundary, such as the
free surface. Waves that travel along the surface are called surface waves. Surface waves are
generally characterized as dispersive; that is, each frequency component travels at a dierent
speed.
There are two common types of surface waves: Rayleigh waves and Love waves. Rayleigh
waves travel along the free surface and give rise to particle motion that is always in a vertical
plane, and elliptical and retrograde with respect to the direction of propagation. Although the
wave motion attenuates rapidly with depth, Rayleigh waves attenuate less rapidly with distance
along the direction of propagation because of their two-dimensional (2-D) character, compared to
P - and S-waves that have three-dimensional (3-D) character. The velocity with which Rayleigh
waves travel is a little less than that of shear waves in the same medium.
In many seismograms recorded on land where the near-surface comprises a low-velocity
layer, Rayleigh waves are observed in the form of a low-velocity, large-amplitude, and lowfrequency wavetrain. These waves are commonly known as ground roll, and they obscure the
useful reection energy on recorded seismograms.
Love waves are surface waves with particle motion that is horizontal and transverse to the
direction of propagation. They occur when there is a low-velocity layer overlying a high-velocity
medium. Because the corresponding particle motion is horizontal, Love waves are usually not
recorded in seismic prospecting by receivers that respond only to vertical displacements.

L.4 Wave Propagation Phenomena Diraction, Reection, and Refraction


A seismic energy source generates a spherical wave in a 3-D earth. Following an explosion, a
spherical cavity is created with its periphery forming a zone of permanent deformation. However,
a short distance away from the cavity, the seismic energy induces an elastic deformation. The
particle motion associated with this deformation can be described by a time-varying function,
commonly known as the source waveform.
It turns out that every point on the spherical wavefront acts like a secondary source that
generates its own spherical wave. This is known as Huygens principle which is the physical
basis for wave propagation. Huygens principle can be used to conveniently describe reection
and refraction at layer boundaries, and diraction from sharp discontinuities in the subsurface.
At suciently large distances away from the source location, the spherical wavefront can
be represented by plane waves traveling at dierent angles from the vertical. Each plane wave
carries along the propagation path the seismic energy contained in the source waveform with a
certain frequency band. The propagation path is known as the raypath which is perpendicular
to the wavefront in isotropic media.

Mathematical Foundation of Elastic Wave Propagation

2013

As the spherical wave propagates outward and away from its source, two things happen
to the source waveform. First, the energy generated by the source is spread over the area of
the spherical wavefront, which increases as the square of the radius of the wavefront. Hence,
the energy per unit area is inversely proportional to the square of the radius, while the wave
amplitude is inversely proportional to the radius. The phenomenon of wave amplitude decaying
with distance from the source is called spherical spreading.
Second, part of the seismic energy is lost along the path of propagation because of intrinsic
attenuation of rocks. This attenuation is in the form of frictional dissipation of energy into heat.
Attenuation mechanisms are still subject to research; nevertheless, a probable cause of loss of
seismic energy from absorption is widely believed to be the presence of pore uids. When set
to motion by the passing wave, uids take out a signicant part of the seismic energy. Wave
attenuation is generally described by an amplitude function that decays exponentially with
distance from the source. The rate of decay depends on the rock type and is described by a
constant factor. The constant fractional energy loss per cycle of seismic waves is highest for
weathered rocks (3 dB/wavelength), and it is about 0.5 dB/wavelength for normal rock types.
Since the energy loss is considered to be constant for a given wavelength, higher frequencies are
attenuated more rapidly than lower frequencies. As the source waveform travels down in the
earth, it gets broader along the propagation path because of gradual loss of higher frequencies.
This means that the earth behaves as a low-pass lter.
Another aspect of wave attenuation is related to the observation made about many of the
earth materials that do not behave as perfectly elastic when subjected to extended loading. This
nonelastic behavior is called viscoelasticity (Fung, 1965), and it requires some modications to
Hookes law given by equation (L-21). Specically, stress components are related to strain components and the rate of change of strain components themselves. When viscosity is introduced
into the wave propagation in solids, we can still identify P - and S-wave types. However, these
waves are now damped and dispersive they are subjected to frequency-dependent attenuation.
It is generally believed that the dispersive nature attributable to viscoelastic wave propagation
is negligible within the range of frequencies used in seismic prospecting.
Seismic waves whether compressional or shear, are subjected to reection and refraction
at layer boundaries with impedance contrast and diraction at a sharp discontinuity. In this
section, we shall briey review diraction phenomenon and discuss reection and refraction of
seismic waves within the context of amplitude variation with oset (AVO) analysis and amplitude
inversion.
When the incident plane wave encounters a sharp discontinuity with the radius of curvature
much less than the seismic wavelength, such as a fault edge, then the energy propagates outward
from that point acting as a Huygens secondary source in the form of a spherical wave. This
phenomenon is known as diraction.
The amplitude of the diracted wave along the hyperbolic traveltime trajectory decays with
distance away from the diractor. For an incident plane compressional plane wave, Sommerfeld
gives the exact solution for the diracted wave amplitude as (Grant and West, 1965)

A(, x, z) = A1 exp i( r + ) ,
(L 54a)

4
where
A1 = A0

1
1
sec ( ) csc ( + ) ,

2
2

(L 54b)

and
A0 is the incident wave amplitude, is the wave velocity, is the angle of incidence,
r = x2 + z 2 , is the angle at which energy diracts, and is the angular frequency. Note that
the solution given by equation (L-54a) is for 2-D geometry and for suciently large distances,
thus the term far-eld solution.
It is interesting to note that a reector can be represented by a series of discrete point
diractors. The response of a series of diractors is simply the superposition of the individual
responses. In the limit, we obtain the traveltime response associated with the reector which is
made up of diractors with an innitesimal separation between them.

2014

Seismic Data Analysis

FIG. L-4. A two-layered 2-D earth model used in deriving the Zoeppritz equations (L-78).

We now review the reection and refraction phenomena. For simplicity, consider a
monochromatic compressional plane wave that impinges at normal incidence upon a at layer
boundary at z = 0. The incident energy is partitioned between a reected and transmitted
compressional plane wave. This special case is discussed in Section 11.2. In the next section, we
shall treat the general case of a compressional plane wave that is incident at an oblique angle to
the interface. Derivation of the reected and transmitted wave functions gets complicated, and
we nd that the reection coecient changes with angle of incidence. Moreover, at non-normal
incidence, the incident compressional-wave energy is partitioned at the interface into four components reected compressional, reected shear, transmitted compressional, and transmitted
shear waves.

L.5 The Zoeppritz Equations


Consider the case of a 2-D medium in (x, z) coordinates that comprises two solid layers separated by a at interface as shown in Figure L-4. We shall adopt the derivation of reected and
transmitted wave amplitudes by Ocer (1958) for this 2-D case between an acoustic layer (water
layer) on top and a solid layer (elastic half space) below to derive the Zoeppritz equations that
describe the reected and refracted P - and S-wave amplitudes for the case of two solid layers,
again separated by a at interface at z = 0.
In many of the wave propagation problems, it is convenient to work with displacement
potentials in lieu of the displacement vector u : (u, w) itself (Ocer, 1958). We shall dene

Mathematical Foundation of Elastic Wave Propagation

2015

two displacement potentials, a compressional-wave potential and a shear-wave potential in


terms of the displacements u and w as
u=


+
x
z

(L 55a)

w=

.
z
x

(L 55b)

and

Choice of the displacement potentials is made on the basis that they satisfy the wave equation
while enabling us to describe the displacement vector u in terms of a compressional component
and a rotational component. To see that the two potentials dened by equations (L-55a,b) satisfy
the wave equation, proceed as follows.
Adapt the dilatation given by equation (L-27) to the geometry of Figure L-4 in (x, z)
coordinates
u w
+
,
x
z

(L 56)



+
.
+

x x
z
z z
x

(L 57)

=
and subsitute equations (L-55a,b)
=

Then, simplify to get the relation


=

2 2
+
.
x2
z 2

(L 58)

Finally, substitute equation (L-58) into the compressional-wave equation (L-34) to obtain
2 2
1 2
+
=
,
x2
z 2
2 t2

(L 59)

where the compressional-wave velocity is given by equation (L-35). Equation (L-59) indicates
that the displacement potential obeys the wave equation.
Now consider the rotational shear component xz given by equation (L-7a). This is the
shear strain that is relevant to the wave motion in (x, z) coordinates of Figure L-4. Substitute
equations (L-55a,b) into equation (L-7a)
xz =

+
2 x z
x
z x
z

(L 60)

and simplify to obtain


xz =

1 2 2
.
+
2 x2
z 2

(L 61)

Finally, substitute equation (L-61) into the shear-wave equation (L-42) to obtain
2 2
1 2
+
=
,
x2
z 2
2 t2

(L 62)

where the shear-wave velocity is given by equation (L-47). Equation (L-62) indicates that the
displacement potential also obeys the wave equation.
The plane-wave solutions (x, z, t) and (x, z, t) to the wave equations (L-59) and (L-62)
in (x, z) coordinates of Figure L-4 are expressed as follows:

sin 0 x + i cos 0 z it
1
1
,

+ A1 exp i sin 1 x i cos 1 z it


1
1

1 = A0 exp i

(L 63a)

2016

Seismic Data Analysis


1 = B1 exp i

sin 1 x i cos 1 z it ,
1
1

(L 63b)

2 = A2 exp i

sin 2 x + i cos 2 z it ,
2
2

(L 63c)

2 = B2 exp i

sin 2 x + i cos 2 z it ,
2
2

(L 63d)

where A0 , A1 , B1 , A2 , and B2 are the incident P -wave, reected P -wave, reected S-wave,
transmitted P -wave, and transmitted S-wave amplitudes, 1 and 1 are the P - and S-wave
velocities in layer 1 (top layer), and 2 and 2 are the P - and S-wave velocities in layer 2
(bottom layer). The angles 0 , 1 , 2 , 1 , and 2 are denoted in Figure L-4. Finally, is the
angular frequency, and the z variable is positive in the downward direction.
Note that the displacement potential 1 given by equation (L-63a) has two components
associated with the incident P -wave and the reected P -wave. To prove that the wave functions
given by equations (L-63a,b,c,d) are solutions to the wave equation, simply substitute these
functions into the respective equations (L-59) and (L-62).
Given the incident P -wave amplitude A0 , our objective is to compute the amplitudes of
the partitioned wave components A1 , B1 , A2 , and B2 . We will require the following four
boundary conditions to be satised at the layer boundary z = 0:
(a)
(b)
(c)
(d)

The
The
The
The

displacement component tangential to the interface is continuous: u1 = u2 .


displacement component normal to the interface is continuous: w1 = w2 .
stress component normal to the interface is continuous: Pzz 1 = Pzz 2 .
stress component tangential to the interface is continuous: Pxz 1 = Pxz 2 .

By way of Snells law that relates the propagation angles


sin 1
sin 2
sin 1
sin 2
sin 0
=
=
=
=
.
1
1
2
1
2

(L 64)

Note that the phase of the wave functions given by equations (L-63a,b,c,d) coincide at the layer
boundary z = 0. With this obervation in mind, refer to equation (L-55a) and consider the rst
boundary condition expressed as
1
1
2
2
+
=
+
.
x
z
x
z

(L 65)

Dierentiate the respective wave functions given by equations (L-63a,b,c,d) and substitute into
equation (L-65), while noting from Snells law that 0 = 1
sin 1
cos 1
sin 2
cos 2
sin 1
A0 +
A1
B1 =
A2 +
B2 .
1
1
1
2
2

(L 66)

Next, refer to equation (L-55b) and consider the second boundary condition expressed as
1
1
2
2

.
z
x
z
x

(L 67)

Dierentiate the respective wave functions given by equations (L-63a,b,c,d) and substitute into
equation (L-67)
cos 1
sin 1
cos 2
sin 2
cos 1
A0
A1
B1 =
A2
B2 .
1
1
1
2
2
Next, refer to equation (L-18c) rewritten below
Pzz = + 2ezz

(L 68)

Mathematical Foundation of Elastic Wave Propagation

2017

and substitute equation (L-56) for and equation (L-3c) for ezz to get
Pzz =

u w
w
+
+ 2
.
x
z
z

(L 69a)

Combine the terms


Pzz = ( + 2)

u
w
+ ,
z
x

(L 69b)

and substitute equations (L-55a,b) for the displacement potentials


Pzz = ( + 2)

+
+
,
z z
x
x x
z

(L 69c)

2
2
2
.
+ 2 2
2
z
x
zx

(L 69d)

2 2
2
2
+

2
+
,
x2
z 2
x2
zx

(L 69e)

and simplify to get


Pzz = ( + 2)
Rewrite equation (L-69d) as
Pzz = ( + 2)

and substitute equations (L-35), (L-47), and (L-59) to get the nal expression for the normal
stress Pzz in terms of the displacement potentials and
Pzz =

2
2
2
2 2
+
.
2
2
t
x
zx

(L 69f )

Now, refer to equation (L-69f) and consider the third boundary condition expressed as
1

2
2
2 1
2 2
2 1
2 2
2 1
2 2

+
=

.
1
2
2
1
2
t2
x2
zx
t2
x2
zx

(L 70)

Dierentiate the respective wave functions given by equations (L-63a,b,c,d) and substitute into
equation (L-70) to get the expression at the boundary z = 0
1 A0 1 A1 + 21

12
2
sin2 0 A0 + 21 12 sin2 1 A1 21 sin 1 cos 1 B1
2
1
1

= 2 A2 + 22

22
sin2 2 A2 + 22 sin 2 cos 2 B2 .
22

(L 71a)

Simplify by combining terms and using the trigonometric relation sin 2 = 2 sin cos to get
12

12
12
2
A
sin

2
sin2 1 A1 sin 21 B1
1
0
12
12

2
2
2
1 2 22 sin2 2 A2 +
sin 22 B2 .
1
2
1

(L 71b)

By way of Snells law given by equation (L-64), note that


12
sin2 1 = sin2 1
12

(L 72a)

22
sin2 2 = sin2 2 .
22

(L 72b)

and

2018

Seismic Data Analysis


Substitute equations (L-72a,b) into equation (L-71b), while noting the trigonometric relation
1 2 sin2 = cos 2
cos 21 A0 cos 21 A1 sin 21 B1 =

2
2
cos 22 A2 +
sin 22 B2 .
1
1

(L 73)

Finally, refer to equation (L-20b) rewritten below as


Pxz = 2exz
and substitute equation (L-5a) for exz to get
Pxz =

w u
+
.
x
z

(L 74a)

Next, substitute equations (L-55a,b) into equation (L-74a) for the displacement potentials
Pxz =

+
+
x z
x
z x
z

(L 74b)

and simplify to get


Pxz = 2

2
2 2

.
+
zx
z 2
x2

(L 74c)

Now, refer to equation (L-74c) and consider the fourth boundary condition expressed as
1 2

2 1
2 1
2 1

+
zx
z 2
x2

= 2 2

2 2
2 2
2 2
.

+
zx
z 2
x2

(L 75)

Dierentiate the respective wave functions given by equations (L-63a,b,c,d) and substitute into
equation (L-75) to get the expression at the boundary z = 0
1

2
2
1
1
sin 0 cos 0 A0 + 2 sin 1 cos 1 A1 2 cos2 1 B1 + 2 sin2 1 B1
2
1
1
1
1

= 2

2
1
1
sin 2 cos 2 A2 2 cos2 2 B2 + 2 sin2 2 B2 .
22
2
2

(L 76a)

Simplify by combining terms and using the trigonometric relation sin 2 = 2 sin cos to get

1
1
1
sin 21 A0 + 2 sin 21 A1 2 (cos2 1 sin2 1 ) B1
2
1
1
1
=

2
2
sin 22 A2 2 (cos2 2 sin2 2 ) B2 .
2
2
2

(L 76b)

Substitute equation (L-47) into equation (L-76b), while noting the trigonometric relations
sin 2 = 2 sin cos and cos 2 = cos2 sin2 to get
sin 21 A0 + sin 21 A1

2 22 12
2 12
12
cos 21 B1 =
sin 22 A2
cos 22 B2 . (L 77)
2
2
2
1
1 1 2
1 12

Equations (L-66), (L-68), (L-73), and (L-77) are the Zoeppritz equations that can be solved
for the four unknowns the reected compressional-wave amplitude A1 , the reected shearwave amplitude B1 , the transmitted compressional-wave amplitude A2 , and the transmitted
shear-wave amplitude B2 .

Mathematical Foundation of Elastic Wave Propagation

2019

Equations (L-66), (L-68), (L-73), and (L-77) are best displayed in matrix form normalized
by the incident-wave amplitude A0 = 1:

1
1
1

sin 1
cos 2
sin 2
cos 1
A1
cos 1

1
2
2

1
1
1
sin 1

cos 1
sin 2
cos 2
B
sin

1
1

1
2
2

=
.

2
2
cos 21

sin
2
cos
2

sin
2
1
2
2 A
cos
2

1
1
1

12
2 22 12
2 12
B2
sin 21
sin 21
2 cos 21
sin 22
cos 22
1
1 12 22
1 12
(L 78)
Given the model parameters and angles of propagation as in Figure L-4, equation (L-78) can
be used to model angle-dependent reection amplitudes. In the next section, we shall discuss
the practical use of an approximate solution of the Zoeppritz equations for P -to-P reection
amplitudes to derive the AVO attributes.

L.6 Prestack Amplitude Inversion


Consider the two-layered earth model in Figure L-4. The Zoeppritz equations (L-78) can be
solved for the amplitudes of the reected and refracted P - and S-wave potentials of equations
(L-63a,b,c,d) A1 , B1 , A2 , and B2 . However, our interest in exploration seismology is largely
the angle dependency of the P -to-P reections given by the coecient A1 . Specically, we wish
to infer or possibly estimate elastic parameters of reservoir rocks from reection amplitudes and
relate these parameters to reservoir uids.
The exact expression for A1 derived from the solution of the Zoeppritz equations (L-78) is
complicated and not intuitive in terms of its practical use for inferring petrophysical properties
of reservoir rocks. Instead, we shall use the approximation provided by Aki and Richards (1980)
as the starting point for deriving a series of practical AVO equations. Now that we only need to
deal with the P -to-P reection amplitude A1 , we shall switch to the conventional notation by
replacing A1 with R() as the angle-dependent reection amplitude for AVO analysis.
By assuming that changes in elastic properties of rocks across the layer boundary are small
and propagation angles are within the subscritical range, the exact expression for R() given by
the Zoeppritz equation can be approximated by (Aki and Richards, 1980)
R() =

1
1 + tan2
2

2
2

1
1 4 2 sin2
4 2 sin2
+

(L 79)

where = (1 + 2 )/2, average P -wave velocity and = (2 1 ), = (1 + 2 )/2, average


S-wave velocity and = 2 1 , = (1 + 2 )/2, average density and = 2 1 , and
= (1 + 2 )/2, average of the incidence and transmission angles for the P -wave (Figure L-4).
Consider the case of N -fold CMP data represented in the domain of angle of incidence. In
discrete form, equation (L-79) can be rewritten as
Ri = ai

+ bi
+ ci
,

(L 80)

where i is the trace index and the coecients ai , bi , and ci are given by
ai =

1
1 + tan2 i ,
2

bi = 4

2
sin2 i ,
2

(L 81a)
(L 81b)

2020

Seismic Data Analysis


and
ci =

2
1
1 4 2 sin2 i .
2

(L 81c)

Note that the reection amplitude Ri is a linear combination of three elastic parameters
fractional change in P -wave velocity /, fractional change in S-wave velocity /, and
fractional change in density /.
We have, for each CMP location and for a specic reection event associated with a layer
boundary, N such equations (L-80) and three unknowns /, /, and /. We have
more equations than unknowns; hence, we encounter yet another example of generalized linear
inversion (GLI) problem. The objective is to determine the three parameters such that the
dierence between the modeled reection amplitudes Ri represented by equation (L-80) and the
actual reection amplitudes Xi is minimum in the least-squares sense (Smith and Gidlow, 1987).
The least-squares error energy S is dened as
N

S=

Ri Xi .

(L 82)

Smith and Gidlow (1987) argue in favor of solving for only two of the three parameters by
making use of the empirical relation between density and P -wave velocity (Gardner et al.,
1974):
= k1/4 ,

(L 83a)

where k is a scalar. This relation holds for most water-saturated rocks. Dierentiate to get
1

=
.

(L 83b)

Now substitute equation (L-83b) into the original Aki-Richards equation (L-80) and combine
the terms with /
R() =

1
2
1
1 + tan2 +
1 4 2 sin2
2
8

4 2 sin2
.

(L 84)

Simplify the algebra to obtain the desired two-parameter model equation for prestack amplitude
inversion
R() =

5 1 2

2
1

4 2 sin2
.
sin2 + tan2
2
8 2
2

(L 85)

Redene the coecients ai and bi and write the discrete form of equation (L-85)
Ri = ai

+ bi
,

(L 86)

where
ai =

5 1 2
1

sin2 i + tan2 i
2
8 2
2

(L 87a)

and
bi = 4

2
sin2 i .
2

(L 87b)

Now substitute equation (L-86) into the error-energy equation (L-82) and expand the terms
N

S=
i

a2i

+ 2ai bi


+ b2i

2 ai

+ bi
Xi + Xi2 .

(L 88)

Mathematical Foundation of Elastic Wave Propagation

2021

Minimization of E with respect to the parameters / and / that we wish to estimate


from prestack amplitude inversion requires that
S
=0
(/)

(L 89a)

S
= 0.
(/)

(L 89b)

and

Compute the partial derivatives using equation (L-88) and substitute into equations (L-89a,b)
to get two simultaneous equations
N

a2i

ai bi
i

ai Xi

(L 90a)

bi X i .

(L 90b)

and
N

b2i

ai bi
i

These equations are put into matrix form

N
i
N
i

a2i

ai bi

=
N 2
b
i
i

N
i

ai bi

N
i

ai Xi

N
i

bi X i

(L 91)

and solved for / and /.


The two parameters / and / represent fractional changes in P - and S-wave velocities; as such, they are related to P - and S-wave reectivities, IP /IP and IS /IS , respectively,
where IP and IS are the P - and S-wave impedances given by
IP =

(L 92a)

IS = .

(L 92b)

and
The fractional change in P -wave impedance is given by
IP =

IP
IP
+
.

(L 93a)

Apply dierentiation to equation (L-92a) and substitute into equation (L-93a), then divide both
sides of the resulting equation by IP to get the expression for P -wave reectivity
IP

=
+
.
IP

(L 93b)

Similarly, the fractional change in S-wave impedance is given by


IS =

IS
IS
+
.

(L 94a)

Apply dierentiation to equation (L-92b) and subsitute into equation (L-94a), then divide both
sides of the resulting equation by IS to get the expression for S-wave reectivity
IS

=
+
.
IS

(L 94b)

Petrophysical interpretation is then made using the crossplot of P - and S-wave reectivities.
Note from the denitions of the coecients ai and bi given by equations (L-87a,b) that you have
to choose a value for the ratio / to compute the two reectivities.

2022

Seismic Data Analysis


An alternative formulation of prestack amplitude inversion is based on transforming the
Aki-Richards equation (L-79) to new variables in terms of Lames constants and (Goodway
et al., 1998). Refer to equation (L-35) and dene a new quantity = + 2. The relationship
between the two sets of variables (, ) and (, , ), by way of equations (L-35) and (L-47), is
given by
= 2

(L 95a)

= 2 .

(L 95b)

and
The fractional changes in and are given by
=

(L 96a)

+
.

(L 96b)

and

Apply dierentiation to equation (L-95a) and substitute into equation (L-96a), then divide
both sides by


=2
+
.

(L 97a)

Similarly, apply dierentiation to equation (L-95b) and substitute into equation (L-96b), then
divide both sides by

=2
+
.

(L 97b)

Now solve equations (L-97a,b) for / and /, respectively, and substitute into the
Aki-Richards equation (L-79)
R() =

1
1 + tan2
4

1
1 + tan2

2
2

1
2 2 sin2
1 4 2 sin2
+ 2 2 sin2
+

(L 98)

Finally, simplify to get the Aki-Richards equation transformed to the new variables / and
/
R() =

1
1 + tan2
4

1
1 tan2
2 2 sin2
+

(L 99)

As for the original Aki-Richards equation (L-79), consider the case of N -fold CMP data
represented in the domain of angle of incidence. In discrete form, equation (L-99) can be rewritten
as

Ri = ai
+ bi
+ ci
,
(L 100)

where i is the trace index and the coecients ai , bi , and ci are given by
ai =

1
1 + tan2 i ,
4

bi = 2

2
sin2 i ,
2

(L 101a)
(L 101b)

Mathematical Foundation of Elastic Wave Propagation

2023

and
1
1 tan2 i .
(L 101c)
4
The reection amplitude Ri in equation (L-100) is a linear combination of three elastic parameters fractional changes /, /, and /. Based on the least-squares minimization
given by equation (L-82), these three parameters can, in principle, be estimated for a specic
event from CMP data.
A variant to the transformation given by equation (L-99) can be formulated in terms of
the P - and S-wave reectivities, IP /IP and IS /IS , respectively. Solve equation (L-93b) for
/ and equation (L-94b) for /, and substitute into the Aki-Richards equation (L-79)
ci =

R() =

1
1 + tan2
2

IP
1
1 + tan2

IP
2

2
2
2
IS

1
4 2 sin2
1 4 2 sin2
+ 4 2 sin2
+

IS

(L 102)

Simplify to get the Aki-Richards equation transformed to the new variables IP /IP and IS /IS
R() =

1
1 + tan2
2

IP
2
1
2
IS

4 2 sin2

tan2 2 2 sin2
.
IP

IS
2

(L 103)

Again, consider the case of N -fold CMP data represented in the domain of angle of incidence.
In discrete form, equation (L-103) can be rewritten as
Ri = ai

IS

IP
+ bi
+ ci
,
IP
IS

(L 104)

where i is the trace index and the coecients ai , bi , and ci are given by
ai =

1
1 + tan2 i ,
2

bi = 4

2
sin2 i ,
2

(L 105a)
(L 105b)

and
2
1
(L 105c)
tan2 2 2 sin2 i .
2

The reection amplitude Ri in equation (L-104) is a linear combination of three parameters


IP /IP , IS /IS , and /. Again, based on the least-squares minimization given by equation
(L-82), these three parameters can be estimated for a specic event from CMP data.
Goodway et al. (1998) implemented a specic form of equation (L-104) to derive the AVO
attributes IP /IP and IS /IS . For a specic value of / = 2 and small angles of incidence,
the third term in equation (L-104) vanishes. The resulting equation then is solved for the P and S-wave reectivities, IP /IP and IS /IS , respectively, using the least-squares solution as
in equation (L-91)

IP N
N 2
N
ai Xi
i
i ai
i ai bi
IP
.

(L 106)
=

IS
N
N
N 2
b
X
a
b
b
i i
i i
i
i
i
i
IS
ci =

2024

Seismic Data Analysis


REFERENCES
Aki, K. I. and Richards, P. G., 1980, Quantitative seismology: W. H. Freeman and Co.
Alam, A. and Millahn, K., 1986, Interactive model-based VSP-CDP transform: Presented at the
Symp. on Practical Aspects of Modeling in Exploration and Development, Kristiansand, Norway.
Alford, R. M., 1986, Shear data in the presence of azimuhtal anisotropy: Dilly, Texas: 56th Ann.
Internat. Mtg., Soc. Expl. Geophys., Expanded Abstracts, 476-479.
Alkhalifah, T., 1996, Transformation to zero oset in transversely isotropic media: Geophysics, 61,
947-963.
Alkhalifah, T. and Larner, K., 1994, Migration error in transversely isotropic media: Geophysics,
59, 1405-1418.
Alkhalifah, T. and Tsvankin, I., 1995, Velocity analysis for transversely isotropic media: Geophysics,
60, 1550-1566.
Alkhalifah, T., Tsvankin, I., Larner, K., and Toldi, J., 1996, Velocity analysis and imaging in
transversely isotropic media: Methodology and a case study: The Leading Edge, 371-378.
Anderson, J. E., Alkhalifah, T., and Tsvankin, I., 1996, Fowler DMO and time migration for
tranversely isotropic media: Geophysics, 61, 835-845.
Anderson, J. E, and Tsvankin, I., 1997, Dip-moveout processing by Fourier transform in anisotropic
media: Geophysics, 62, 1260-1269.
Ata, E. and Michelena, R., 1995, Mapping distribution of fractures in a reservoir with P S -converted
waves: The Leading Edge, 664-673.
Barr, F. J. and Sanders, J. I., 1989, Attenuation of water-column reverberations using pressure
and velocity detectors in a water-column cable: 59th Ann. Internat. Mtg., Soc. Expl. Geophys.,
Expanded Abstracts, 653-656.
Bernitsas, N., 1990, Curvature eect of a reecting surface for arbitrary oset and azimuth: 60th
Ann. Internat. Mtg., Soc. Expl. Geophys., Expanded Abstracts, 1053-1055.
Bortfeld, R., 1961, Approximation to the reection and transmission coecients of plane longitudinal and transverse waves: Geophys. Prosp., 9, 485-502.
Bracewell, R., 1965, The Fourier transform and its applications: McGraw-Hill Book Co.
Caldwell, J., 1999, Marine multicomponent seismology: The Leading Edge, 1274-1282.
Cassell, B., Alam, A., and Millahn, K., 1984, Model-based interactive VSP-CDP transformation:
Presented at the 54th Ann. Internat. Soc. Expl. Geophys.
Castagna, J. P., 1993, AVO analysis tutorial and review: in Oset-dependent reectivity
theory and practice, Soc. Expl. Geophys.
Castagna, J. P., Bazle, M. L., and Eastwood, R. L., 1985, Relationship between compressional-wave
and shear-wave velocities in elastic silicate rocks: Geophysics, 50, 571-581.
Cooke, D. A. and Schneider, W. A., 1983, Generalized linear inversion of reection seismic data:
Geophysics, 48, 665-676.
Daley, P. F. and Hron, F., 1979, Reection and transmission coecients for seismic waves in
ellipsoidally anisotropic media: Geophysics, 44, 27-38.
Demirbag, E. and Coruh, C., 1988, Inversion of Zoeppritz equations and their approximations:
58th Ann. Internat. Mtg., Soc. Expl. Geophys., Expanded Abstracts, 1199-1203.
den Rooijen, H. P. G. M., 1991, Stacking og P SV reection data using dip moveout: Geophys.
Prosp., 39, 585-598.
Dillon, P. B. and Thomson, R. C., 1983, Image reconstruction for oset source VSP surveys:
Presented at the 45th Ann. Eur. Assoc. Expl. Geophys. Mtg.
Dragoset, W. and Barr, F. J., 1994, Ocean-bottom cable dual-sensor scaling: 64th Ann. Internat.
Mtg., Soc. Expl. Geophys., Expanded Abstracts, 653-656.
Eaton, D. W. S. and Lawton, D. C., 1992, P -SV stacking charts and binning periodicity: Geophysics, 57, 745-748.
Eastwood, R. L. and Castagna, J. P., 1983, Basis for interpretation of vP /vS ratios in complex
lithologies: 24th Ann. Meeting of Logging Symp., Soc. Prof. Well Log Analysts.

Mathematical Foundation of Elastic Wave Propagation


Ecker, C., Lumley, D. E., Tura, A., Kempner, W., Klosnky, L., 1999, Estimating separate steam
thickness and temperature maps from 4-D seismic data: An example from San Joaquin Valley,
California: 69th Ann. Internat. Mtg., Soc. Expl. Geophys., Expanded Abstracts, 2032 - 2034.
Fatti, J. L., Smith, G. C., Vail, P. J., Strauss, P. J., and Levitt, P. R., 1994. Detection of gas
in sandstone reservoirs using AVO analysis: A 3-D seismic case history using the Geostack
technique: Geophysics, 59, 1362-1376.
Fromm, G., Krey, T., and Wiest, B., 1985, Static and dynamic corrections, in Dohr, G., Ed., Seismic
Shear Waves: Handbook of Geophysical Exploration, vol. 15a: Geophysical Press, 191-225.
Fung, Y. C., 1965, Foundations of solid mechanics: Prentice-Hall.
Gaiser, J. E., 1998, Compensating OBC data for variations in geophone coupling: 68th Ann. Internat. Mtg., Soc. Expl. Geophys., New Orleans, Expanded Abstracts, 1429-1432.
Gaiser, J. E., 1999a, Applications of vector coordinate systems of converted waves obtained by
multicomponent 3-D data: Presented at the Oshore Technology Conference, Paper No. 10985.
Gaiser, J. E., 1999b, Applications of vector coordinate systems of 3-D converted-wave data: The
Leading Edge, 1290-1300.
Gaiser, J. E. and Jackson, A. R., 1998, 3-D P S -wave midpoint transformation in variable media:
68th Ann. Internat. Mtg., Soc. Expl. Geophys., Expanded Abstracts, 1417-1420.
Gardner, G. H. F., Gardner, L. W., and Gregory, A. R., 1974, Formation velocity and density
The diagnostic basis for stratigraphic traps: Geophysics, 39, 770-780.
Grant, F. S. and West, G. F., 1965, Interpretation theory in applied geophysics: McGraw-Hill Book
Co.
Goodway, B., Chen, T., and Downton, J., 1998, AVO and prestack inversion: Presented Ann. Mtg.
Can. Soc. of Expl. Geophys.
Grechka, V. and Tsvankin, I., 1997, Inversion of nonhyperbolic moveout in transversely isotropic
media: 67th Ann. Internat. Mtg., Soc. Expl. Geophys., Expanded Abstracts, 1685-1688.
Haase, A. B., 1998 (November), Nonhyperbolic moveout in Plains data and the anisotropy question:
The Recorder, Can. Soc. Expl. Geophys., 20-34.
Hale, D. and Artley, C., 1992, Squeezing dip-moveout for depth-variable velocity: Geophysics, 58,
257-264.
Hardage, B. A., 1983, Vertical seismic proling: Geophysical Press.
Hilterman, F. J., 1970, Three-dimensional seismic modeling: Geophysics, 35, 1020-1037.
Hilterman, F. J., 1975, Amplitudes of seismic waves A quick look: Geophysics, 40, 745-762.
Hilterman, F. J., 1982, Interpretive lessons from three-dimensional modeling: Geophysics, 47, 7848012.
Hilterman, F. J., 1983, Seismic lithology: presented as a continuing education course at the 53rd
Ann. Internat. Mtg., Soc. Expl. Geophys.
Kallweit, R. S. and Wood, L. C., 1982, The limits of resolution of zero-phase wavelets: Geophysics,
47, 1035-1046.
Kitchenside, P. W., 1993, 2-D aniostropic migration in the space-frequency domain: J. of Seis.
Expl., 2, 7-22.
Koefoed, O., 1955, On the eect of Poissions ratio of rock strata on the reection coecients of
plane waves: Geophys. Prosp., 3, 381-387.
Larner, K. L., 1993, Dip-moveout error in transversely isotropic media with linear velocity variation
in depth: Geophysics, 58, 1442-1453.
Lass, H., 1950, Vector and tensor analysis: McGraw-Hill Book Co.
Levin, F. K., 1971, Apparent velocity from dipping interface reections: Geophysics, 36, 510-516.
Levin, F. K., 1990, Reection from a dipping plane transversely isotropic solid: Geophysics, 55,
851-855.
Li, Xiang-Yang and Yuan, Jianxin, 1999, Developing and exploiting eective techniques to overcome
the diculties of 4-C seismic reservoir monitoring in deep water: Edinburgh Anisotropy Project
Ann. Report, 129-155.
Lindseth, R., 1979, Synthetic sonic logs a process for stratigraphic interpretation: Geophysics,
44, 3-26.

2025

2026

Seismic Data Analysis


Lumley, D. E., 1995a, Seismic time-lapse monitoring of subsurface uid ow: Ph. D. thesis, Stanford
University.
Lumley, D. E., 1995b, 4-D seismic monitoring of an active steamood: 65th Ann. Internat. Mtg.,
Soc. Expl. Geophys., Expanded Abstracts, 203-206.
Lumley, D. E., Nur, A., Strandenes, S., Dvorkin, J., and Packwood, J., 1994, Seismic monitoring of
oil production: A feasibility study: 64th Ann. Internat. Mtg., Soc. Expl. Geophys., Expanded
Abstracts, 319-322.
Martin, M. A. and Davis, T. L., 1987, Shear-wave birefringence: A new tool for evaluating fractured
reservoirs: The Leading Edge, 22-28.
MacLeod, M. K., Hanson, R. Bell, R. C., and McHugo, S., 1999, The Alba Field ocean bottom
cable seismic survey: Impact on development: The Leading Edge, 1306-1312.
Miller, S. L. M. and Stewart, R. R., 1999, Eects of lithology, porosity and shaliness on P - and
S -wave velocities from sonic logs: SEG Continuing Education Class Notes.
Oldenburg, D. W., Scheuer, T., and Levy, S., 1983, Recovery of the acoustic impedance from
reection seismograms: Geophysics, 48, 1318-1337.
Ocer, C. B., 1958, Introduction to the theory of sound transmission with application to the ocean:
McGraw-Hill Book Co.
Ostrander, W. J., 1984, Plane-wave reection coecients for gas sands at nonnormal angles of
incidence: Geophysics, 49, 1637-1648.
Paenholz, J. and Barr, F. J., 1995, An improved method for deriving water-bottom reectivities for
processing dual-sensor ocean-bottom cable data: 65th Ann. Internat. Mtg., Soc. Expl. Geophys.,
Expanded Abstracts, 987-990.
Pickett, G. R., 1963, Acoustic character logs and their applications in formation evaluation: J. Can.
Petr. Tech., 15, 659-667.
Probert, T., Hoare, R., Ronen, S., Godfrey, R. J., Pope, D., and Kommedal, J., 1999 (November),
Imaging through gas using 4-C 3-D seismic data: a case study from Lomond Field: Petr. Expl.
Soc. of Great Britain Newsletter.
Richards, T. C., 1961, Motion of the ground on arrival of reected longitudinal and transverse
waves at wide-angle reection distances: Geophysics, 27, 277-297.
Ricker, N., 1953, Wavelet contraction, wavelet expansion and the control of seismic resolution:
Geophysics, 18, 769-792.
Rickett, J. and Lumley, D. E., 1998, A cross-equalization processing ow for o-the-shelf 4-D
seismic data: 68th Ann. Internat. Mtg., Soc. Expl. Geophys., Expanded Abstracts, 16-19.
Ristow, D. and Ruehl, T., 1997, Migration in transversely isotropic media using implicit operators:
67th Ann. Internat. Mtg., Soc. Expl. Geophys., Expanded Abstracts, 1699-1702.
Rueger, A., 1997, P -wave reection coecients for transversely isotropic models with vertical and
horizontal axis of symmetry: Geophysics, 62, 713-722.
Rutherford, S. R. and Williams, R. H., 1989, Amplitude-versus-oset variations in gas sands:
Geophysics, 54, 680-688.
Sheri, R. E., 1991, Encyclopedic dictionary of exploration geophysics: Soc. Expl. Geophys.
Shuey, R. T., 1985, A simplication of the Zoeppritz equations: Geophysics, 50, 609-614.
Shuey, R. T., Banik, N. C., and Lerche, I., 1984, Amplitude from curved reectors: 54th Ann.
Internat. Mtg., Soc. Expl. Geophys., Expanded Abstracts, 664-665.
Smith, G. C. and Gidlow, P. M., 1987, Weighted stacking for rock property estimation and detection
of gas: Geophys. Prosp., 35, 993-1014.
Soubaras, R., 1996, Ocean-bottom hydrophone and geophone processing: 66th Ann. Internat. Mtg.,
Soc. Expl. Geophys., Expanded Abstracts, 24-27.
Spratt, R. S., Goins, N. R., and Fitch, T. J., 1984, Pseudo-Shear The analysis of AVO: in
Oset-dependent reectivity theory and practice: Soc. Expl. Geophys.
Stolt, R. H. and Benson, A. K., 1986, Seismic migration theory and practice: Geophysical Press,
London-Amsterdam.
Taner, M. T., 1978, Complex seismic trace analysis: Geophysics, 44, 1041-1063.
Taner, M. T. and Koehler, F., 1981, Surface-consistent corrections: Geophysics, 46, 17-21.

Mathematical Foundation of Elastic Wave Propagation


Tessmer, G. and Behle, A., 1988, Common reection point data stacking technique for converted
waves: Geophys. Prosp., 36, 671-688.
Theilen, F., Ayres, A., and Lange, G., 1997, Phyiscal properties of near-surface marine sediments:
59th EAGE Mtg., Extended Abstracts.
Thomsen, L., 1986, Weak elastic anisotropy: Geophysics, 51, 1954-1966.
Thomsen, L., 1998, Converted-wave reection seismology over anisotropic, inhomogeneous media:
68th Ann. Internat. Mtg., Soc. Expl. Geophys., Expanded Abstracts, 2048-2051.
Tsvankin, I., 1995, Normal moveout from dipping reectors in anisotropic media: Geophysics, 60,
268-284.
Tsvankin, I., 1996, P -wave signatures and notation fro transversely isotropic media: An overview:
Geophysics, 61, 467-483.
Tsvankin, I. and Thomsen, L., 1994, Nonhyperbolic reection moveout in aniostropic media: Geophysics, 59, 1290-1304.
Tura, A. and Lumley, D. E., 1999, Estimating pressure and saturation changes from time-lapse
AVO data: 69th Ann. Internat. Mtg., Soc. Expl. Geophys., Expanded Abstracts, 1655-1658.
Uren, N. F., Gardner, G. H. F. and McDonald, J. A., 1990a, The migrators equation for anisotropic
media: Geophysics, 55, 1429-1434.
Uren, N. F., Gardner, G. H. F. and McDonald, J. A., 1990b, Dip moveout in anisotropic media:
Geophysics, 55, 863-867.
Uren, N. F., Gardner, G. H. F. and McDonald, J. A., 1990c, Normal moveout in anisotropic media:
Geophysics, 55, 1634-1636.
Verwest, B. J., 1989, Seismic migration in elliptically anisotropic media: Geophys. Prosp., 37,
149-166.
Widess, M. B., 1973, How thin is a thin bed?: Geophysics, 38, 1176-1180.
Wiggins, R., Kenny, G. S., and McClure, C. D., 1984, A method for determining and displaying
the shear-wave reectivities of a geologic formation: European patent application (Mobil Oil
Corporation).
Winterstein, D. F., 1990, Velocity anisotropy terminology for geophysicists: Geophysics, 55, 10701088.
Wyatt, K. and Wyatt, S.B., 1981, Determination of subsurface structural information using the
vertical seismic prole: Presented at the 51st Ann. Internat. Mtg., Soc. Expl. Geophys.
Yuan, J. and Li. X-Y., 1997, Converted-wave CCP binning and velocity analysis: in Processing
three-component seaoor seismic data, Edinborough University.
Zhang, Y. and Robinson, E. A., 1992, Stacking P -SV converted wave data with raypath velocity:
62nd Ann. Internat. Mtg., Soc. Expl. Geophys., Expanded Abstracts, 1214-1217.
Zhang, Y., 1996, Nonhyperbolic converted wave velocity analysis and normal moveout: 66th Ann.
Internat. Mtg., Soc. Expl. Geophys., Expanded Abstracts, 1555-1558.
Zhu, X., Altan, S. and Li, J., 1999, Recent advances in multicomponent processing: The Leading
Edge, 1283-1288.

2027

Você também pode gostar