Professional Documents
Culture Documents
AVO Theory 1
Contents
Introduction
Amplitude versus offset (AVO) analysis is a relatively new seismic technique used very
successfully to predict hydrocarbon (especially gas) and lithologies for oil exploration and
reservoir characterization.
Using the AVO technique seismic amplitudes and their variation as a function of offset are
analyzed. This information is then used to predict lithology and fluid content.
The basic concept behind the AVO technique is the analysis of the reflection process. We
take advantage of the fact that the measured amplitude is related to the strength of the
reflection (reflection coefficient), and that the reflection coefficient depends on three
parameters:
1. Change in P-wave velocity across the interface.
2. Change in S-wave velocity across the interface.
3. Change in density across the interface.
This introduction to AVO theory is divided into three sections. In the first section we discuss
basic rock physic concepts showing you how fluid changes affect elastic parameters. In the
second section there is a discussion of basic reflection theory and you gain an understanding
of how amplitude varies with offset. In the third section you are guided through AVO inversion
which involves converting seismic data to AVO attributes. AVO attributes are seismic data
volumes which are proportional to the change in a seismic - rock parameter such as Vp, Vs,
etc.
Overview
This section consists of a review of terms used in rock physics, and an overview of basic
elastic parameters and rock physics formulas.
Fig. 1 displays a thin section of a rock. Note that the rock is not homogenous but rather a
collection of grains made of different minerals of varying shapes and sizes. The rock is held
together by cement. In between the grains is pore space which is usually filled with fluid.
Basic Terms
Density ρ is the mass of the rock per unit of volume. The bulk density of the rock is affected
by the composition of the different minerals, the porosity of the rock, and the type of fluids
which fill the pore space.
Porosity φ is the percentage of pore space per unit volume. Note that porosity can never
reach 100% which would mean that the rock does not contain any minerals only pores. There
is a point - called critical porosity - φ c where the rock is considered to still be a rock. Above
critical porosity the rock becomes a suspension. See Fig. 2.
Saturation ( S ) defines the quantity of a specific fluid in the pore space. Saturation requires
the specification of fluid type. For example, water saturation S w is the percentage of brine
compared to hydrocarbon within the pore space. 40% S w means that the fluid is composed
of 40% water and 60% hydrocarbons.
The overall elastic rock property is determined by the properties of the rock matrix, the
porosity, and the composition of the fluids that fill the pore space.
Elasticity: The wave propagation process which occurs during a seismic experiment is
controlled by the elastic properties of the rock - meaning how the rock deforms as a result of
the force which is applied to it. Intuitively it is obvious that soft rock reacts to the strain
differently than does a stiffer rock.
Elasticity theory deals with the deformation caused by the application of stress to a
substance.
Hook’s Law: States that there is a linear relationship between stress and strain (between the
force applied and the amount of deformation).
σ = C⋅e (EQ 1)
C is a constant relation. Strain( e ) and stress( σ ) are tensors and C is a matrix (tensor) of
constants (with 81 coefficients) which define the fundamental elastic properties of the rock.
For isotrophic material, the 81 coefficients of matrix C are reduced to two independent elastic
parameters which characterize the elastic properties of the rock. Some combinations of the
two independent parameters are called the Elastic Moduli and can be measured in laboratory
experiments.
∆L
σ l = E ------- (EQ 2)
L
See Fig. 3.
∆Y
σ s = µ ------- . (EQ 3)
X
See Fig. 4.
∆V
P = K -------- , (EQ 4)
V
where:
P = hydrostatic pressure
∆V
-------- = relative change in volume
V
See Fig. 5.
Bulk modulus is the elastic modulus used mostly in AVO. The table in Fig. 6 lists several
typical values of Bulk modulus for some rocks and fluids. The units are in 1010 Dynes/cm2.
Note that there is a large difference in Bulk modulus between solids and fluids. It is obvious
that if the solid is stiff, it will react differently to the application of volumetric stress.
Lame Constant λ is another elastic modulus that is used very often. It is not a property
which is directly measured in the laboratory but is defined from other elastic moduli.
K = λ + 2 ⁄ 3µ (EQ 5)
Wave Propagation Velocities: Two basic waves propagate in the subsurface: Pressure
wave (P-wave) and Shear wave (S-wave). Note that Shear waves do not propagate through
fluids.
P-waves and S-waves propagate with different velocities, and those velocities are related to
the elastic properties of the rocks.
k + 3 ⁄ 4µ
Vp = ----------------------- (EQ 6)
ρ
µ
Vs = --- (EQ 7)
ρ
Note that the S-wave velocity is affected by density and the Shear modulus. P-wave velocity is
affected by two elastic moduli, the Bulk modulus and the Shear modulus. In other words, the
Bulk modulus only affects the P-wave.
The relationship between P-wave velocity and S-wave velocity is a fundamental parameter in
AVO analysis.
Exx
σ = – --------- (EQ 8)
Ezz
2
Vp
------- – 2
Vs
σ = ------------------------------ (EQ 9)
2
2 Vp ------- – 1
Vs
Vp
This means that Poisson’s ratio is a measure of ------- . Note that Poisson’s ratio varies between
Vs
0 and 0.5.
σ = 0 when ------- =
Vp
2
Vs
σ = 0.5 when Vs = 0 .
Is = Vs ⋅ ρ (EQ 10)
Impedance plays a major role in Normal Incidence reflectivity. At Normal Incidence, the
reflection from an interface between two layers, with Ip1 representing impedance of the top
layer, and Ip2, the impedance of the bottom layer, is given by:
I p 2 – I p 1 ∆Ip
R = ------------------------- = --------- - relative changes in impedance. (EQ 11)
I p 2 + I p 1 Ip
This is the reason why impedance can be directly measured from normal incidence seismic
data.
The seismic experiment is controlled by the velocity of the waves in the subsurface. Because
of the large wavelength of seismic waves in oil exploration (in the order of meters), the
velocity and density that we can measure from the seismic data reflects some average
property of the rock. It is a combination of the parameters of the rock matrix (the grains) and
the fluids that fill the pore spaces (which are in the scale of millimeters).
1 1 – φ S w φ ( 1 – S w )φ
------- = ------------ + ---------- + ------------------------ (EQ 12)
Vb VM Vw V HC
where:
φ = Porosity (%)
V w = Velocity of water
V HC = Velocity of hydrocarbons
Wyllie’s equation defines the averaging law for velocities as averaging the inverse of the
velocities.
Wyllie’s Equation for Density defines the averaging law for densities. It states
that the bulk density of the rock is the average of the density of its components.
ρb = ρ M ( 1 – φ ) + ρ w S w φ + ρ HC ( 1 – S w )φ (EQ 13)
where:
ρ w = Density of water
φ = Porosity
An analysis of Wyllie’s time average equation shows that the velocity decreases as water is
replaced with hydrocarbons. This is because the Bulk modulus of oil is lower than water and
the Bulk modulus of gas is lower than oil. Therefore the velocity of oil is lower than that of
water.
The diagram in Fig. 7 is a plot of Vp as a function of water saturation. There is one graph for
oil and another for gas showing the decrease of P-wave velocity with the increase in
hydrocarbons. The diagram in Fig. 7 is calculated using Wyllie’s time average equation.
The difference in P-wave velocity between water and hydrocarbons, due to the difference in
Bulk modulus, is the basic phenomena that enables AVO analysis.
Laboratory measurements of rocks show that Wyllie’s time-average equation does not
correctly describe the effect of S w on the P-wave velocity. The dashed line in Fig. 7 plots the
observed velocity showing the discrepancy between observed velocity and the prediction
from Wyllie’s equation.
2
1 – Kb ⁄ KM
2 (EQ 14)
V p ⋅ ρ b = K b + 4 ⁄ 3 µ b + --------------------------------------------------------------------------------
1 – φ – Kb ⁄ KM ⁄ KM + φ ⁄ k f
where:
Vp = P-wave velocity
K = Bulk modulus
µ = Shear modulus
ρ = Density
b, M , f = Bulk, rock matrix, and fluid
1 Sw 1 – Sw
-------- = -------- + ---------------- (EQ 15)
Kf K w K HC
k + 3 ⁄ 4µ
Vp = ----------------------- (EQ 16)
ρ
When porosity is equal to 0, the last term in the Gassman equation is reduced to 0. This
means that the Gassman equation can be viewed as an extension of the formula that defined
P-wave velocity in order to incorporate the fluid effects. The Gassman equation predicts that
upon an introduction of a very small amount of gas, the bulk P-wave velocity will drop.
Increasing the amount of gas thereafter has only a small effect on the velocity. See the
dashed line in Fig. 7.
This phenomena demonstrates that AVO technique can be useful in detecting hydrocarbons,
but is not indicative enough to detect the amount of gas. We expect to see AVO anomalies
even for low gas saturations. Fig. 8 summarizes the prediction of the Gassman equations.
Fig. 8 also illustrates the basis of AVO technique by showing the effects of gas on rock
velocity. Note that when oil is present there is a somewhat different situation because there is
not such a large drop in velocity. This is why it is more difficult to predict oil with AVO
compared to gas although in some cases it is possible.
Note that shear-wave velocity increases slightly with gas saturation. This is due to the effect
of density and is not related to Bulk modulus. It is explained as follows:
µ
Vs = --- (EQ 17)
ρ
µ - the Shear modulus is not affected at all by the fluids, but the density is. As gas saturation
increases, density is reduced. Since density is in the denominator, it causes Vs to increase
slightly.
Poisson’s ratio, which is a function of Vp/Vs, responds mainly to the change in fluids. The
value of Vp responds mainly to the change in porosity and less to the change in pore fluids.
Vp = A ⋅ Vs + B (EQ 18)
where A and B are constants specific to each case.
Gardner’s Relation is another useful relationship between P-wave velocity and density,
defined by Gardner from experimental data. The Gardner Relation is given by:
Overview
AVO technique is the analysis of the variation of the amplitude as a function of offset. The
amplitude varies with offset because the reflection coefficient varies when the angle of
incidence of the wave at the interface varies.
ρ 2 Vp 2 – ρ 1 Vp 1 ∆I p
R NI = -------------------------------------- = --------- (EQ 21)
ρ 2 Vp 2 + ρ 1 Vp 1 I p
When the wave arrives at the interface at non-normal incidence, some of the P wave energy
is converted to shear. Fig. 9 describes this phenomena.
AVO Analysis
AVO analysis is the analysis of the variations in the reflection coefficient. These variations are
governed by the contrast in P-wave velocity and S-wave velocity at the interface. As previously
stated, when there is gas in the layer, Vp drops whereas Vs does not change. This means that
Vp/Vs is anomalous, and we hope to see the effect of this anomaly in the reflection pattern.
In principle AVO analysis should measure amplitude variations with angle of incidence.
However, amplitude is measured with offset because usually as offset increases, the angle of
incidence increases. We also assume that amplitude of the seismic data is proportional to the
reflection coefficients.
Zoeppritz Equations relate the reflection coefficients to the angle of incidence, ∆Vp , ∆Vs ,
and ∆ρ . The Zoeppritz equations are quite complex. This may be the reason why AVO is a
relatively new technique. AVO theory evolved only after approximations to the Zoeppritz
equations were developed. See “Appendix A: Zoeppritz Equations” on page 40.
The basic approximation to the Zoeppritz equations was given by Aki & Richards. See Fig. 10
and Fig. 11.
Note that:
Fig. 11
∆V ∆V
∆ρ p s
R ( θ ) ≈ a ------ + b ----------- + c ---------- (EQ 22)
ρ V V
p s
The coefficients a , b , and c are a function of the angle of incidence θ and the background
Vp
trend ------- . Equation 22 is an approximation to the Zoeppritz equations. It assumes that:
Vs
• It is a plane wave.
• There is P wave to P wave reflection.
• There are only precritical reflections.
• There are only small changes in elastic parameters across the interface.
Rearranging Equation 22 we get as shown in Fig. 12.
Fig. 13
The Hilterman approximation: Hilterman suggested another way to rearrange this formula
1
assuming small angles and that σ ≈ --- .
3
9
R ( θ ) ≈ Rp ⋅ cos 2 θ + --- ∆ σ sin 2 θ (EQ 25)
4
This formula shows that reflectivity at small angles is dominated by P-wave velocity and large
angle reflectivity is dominated by ∆σ (change in Poisson’s Ratio).
9
∆σ = --- ( Rp + G ) (EQ 26)
4
We assume that for small angles, the offset is small and for large angles the offset is large.
See Fig. 14.
During AVO inversion we actually convert the offset to angle of incidence using one of the
ray tracing processes. See “Section Three: AVO Inversion” on page 25.
AVO Anomalies
The basic model for AVO anomaly can be illustrated using a gas sand layer between two
shales. See Fig. 15.
Poisson’s ratio for the gas sand is much smaller than for shale, and the reflection
coefficients change rapidly with the angle of incidence. The resulting amplitude on an
offset gather is negative reflection of near offset becoming more negative at the far offset,
which is manifested with an absolute value of amplitude increase at large offsets. This is
shown in Fig. 15.
• Class III is the classic Gulf of Mexico AVO anomaly where gas sands have low impedance
compared to shale. The reflection coefficient of normal incidence is negative and becomes
more negative at a large offset. Fig. 17 shows that amplitude increases with offset for class
III AVO anomalies.
Fig. 17
Fig. 18
• Class I is a high impedance sand anomaly. It has positive normal incidence reflectivity
which becomes smaller at larger offsets as shown in Fig. 19.
Fig. 19
Bright spots and dim spots: Before AVO analysis was developed, interpreters often used
the bright spots on stack sections as gas indicators. The bright spots were large amplitude
reflections on stack sections which often coincided with gas. The bright spots are associated
with class III AVO anomalies. When you stack them you get high stack amplitudes.
Overview
AVO inversion is the basic step in AVO processing. In this step we attempt to convert the
seismic data to different reflectivities which have clear physical meanings. In AVO inversion
we take the seismic pre-stack data, which represents the reflection coefficient as a function of
angle of incidence and, using the Zoeppritz equations, we try to convert this data to
reflectivity. Reflectivity means relative change in a rock parameter.
∆Vp
• ---------- - relative change in P-wave velocity
Vp
∆Vs
• ---------- - relative change in S-wave velocity
Vs
∆ρ
• ------ - relative change in density
ρ
∆V ∆V
∆ρ s p
R ( θ ) ≈ a ------ + b ---------- + c ----------- , (EQ 27)
ρ V V
s p
and given:
R ( xi ) : Reflection Coefficient (Amplitude) at N distinct offsets X i .
θ angle of incidence at offset X .
i i
∆ρ ∆Vp ∆Vs
Calculate: ------ ; ---------- , ---------- .
ρ Vp Vs
Calculating Angle of Incidence: To perform AVO inversion we first need to know the angle
of incidence for all gathers, offsets, and time (or depth) samples.
Given the velocity model and the source receiver location, we can determine the angle of
incidence using a ray tracing procedure.
The Least Square Formulation for AVO Inversion: The second step in AVO inversion
involves picking all amplitude values (of a given CMP gather) for all offsets for each time
sample. See Fig. 21.
We analyze the amplitude as a function of angle of incidence as illustrated in Fig. 22. Note
that we convert each offset to its corresponding angle using the ray tracing procedure.
Using the least squares formulation we then fit a Zoeppritz approximation curve.This means
that we find the best Zoeppritz curve that fits the data points in a least squares sense.
This process is repeated for all CMPs and for all samples. The results are reflectivity data
cubes, which are also called AVO attributes.
In theory, a straight forward solution of all three parameters is possible. In practice however,
the density term is relatively small compared with velocity. In addition, the density affects the
very far offsets beyond 30o. Unfortunately in most cases far offset data which contains the
density information does not exist. Other factors such as noise and imperfect amplitude
recovery makes inversion for density instable and causes instability in the velocity reflectivity
estimation as well. To stabilize the inversion we usually invert for two-terms rather than the
three-terms that are involved in the Zoeppritz equation.
∆ρ ∆Vp ∆Vs
In the Aki & Richards formula: R ( θ ) ≈ a ------ + b ---------- + c ----------
ρ Vp Vs
∆ρ ∆Vp ∆Vp ∆V s
we substitute: ------ = γ ⋅ ---------- and solve for: R ( θ ) ≈ b ---------- + c ----------
ρ Vp Vp Vs
The basic attributes calculated in AVO inversion using the two-term Aki & Richards
approximation are P-wave velocity reflectivity and S-wave velocity reflectivity. We can
calculate other useful attributes which are a function of the basic two attributes.
Fig. 24
Vp ∆q
If q = ------- is Pseudo Poisson’s ratio, then Pseudo Poisson’s reflectivity ------ is given by:
Vs q
∆q ∆Vp ∆Vs
------ = ---------- – ---------- (EQ 28)
q Vp Vs
Elastic Moduli Reflectivities: Instead of inverting for impedances or velocities we can invert
directly for elastic moduli. (Note that P and S velocities are defined in terms of elastic moduli.
λ + 2µ µ
Vp = ---------------- Vs = --- .)
ρ ρ
AVO Inversion using the Shuey Approximation: The attributes defined so far are
generated by using the Aki & Richards approximation in a process of fitting the curve to the
AVO data.
The inversion process which is based on Shuey’s approximation, is very similar to the A&R
process described above. Here, instead of fitting an A&R curve, we fit the Shuey curve.
Note that the normal incidence reflectivity is identical to P-wave impedance reflectivity.
∆σ
-------------------- = NI + G (EQ 30)
2
(1 – σ)
Creating angle stacks is a simplified form of AVO inversion which is qualitative rather than
quantitative but is quite common and rather useful.
How many attributes do you need? You have already noticed that there are many AVO
attributes. Users often wonder which attribute to use. The answer is that there are only two
∆Vp ∆Vs
independent AVO attributes (when using two term inversion). ---------- and ---------- are examples.
Vp Vs
NI and G are another example. Other attributes are combinations which can be useful. The
significant factors in attribute calculation which can lead to a difference in the results are:
Ideally we search for attributes which can be useful for indicating hydrocarbons by separating
fluids from lithology. Another criteria which affects our choice of attribute is if we intend to use
this attribute in acoustic impedance inversion.
Fig. 25
Fig. 25 displays five different AVO attributes from the same dataset. Note the differences
between the near and far angle stacks. The oil sands are seen at far angle stacks but not at
the near. In the conventional stack only a hint of the AVO anomaly can be seen. Note that the
P-wave reflectivity is very similar to the near angle stacks and that the fluid factor attribute
shows the anomaly best, as expected.
You may want to use other AVO attributes. The Probe application provides you with a
calculator which enables you to create your own attributes by combining existing attributes
with basic functions.
Consider a two layer model. The bottom layer has the following parameters:
Vp = 20,000 m/sec
Vs = 10, 000 m/sec
ρ = 2.65 g/cc
Consider five cases where the top layer parameters for each case are described in the table
in Fig. 26.
Fig. 26
The reflection coefficient R ( θ ) is drawn in Fig. 27 for the five cases; each case in its own
color.
We see that the shape of this curve (calculated with the true Zoeppritz curve) is similar in all
cases. In each case critical angle (where curves end and R = 1) is reached elsewhere as
expected. Also the value of R is different for each case, but the shape is similar. It decreases
at the beginning and at some points it curves and starts to increase.
The three-term inversion approximates this behavior quite well. In fact with three-term
inversion the Shuey curve and the Aki & Richards curves are identical.
With two-term inversion the Aki & Richards curve is different from the Shuey curve and
sometimes behaves better, especially at larger angles. With two-term Shuey, the curves
2
always increase becauseG sin θ is always positive.
R ( θ ) = NI + G sin 2 θ
shuey curve
Amplitude
true Zoeppritz curve
A&R Approximation
Fig. 28
The dark blue curve is a true Zoeppritz curve. (Note that this data has large angles over 50o.
Observe the bending of the curve.) The light blue curve is the A&R approximation and the
green curve is the Shuey curve. A&R works much better with these large angles. For smaller
angles (up to 30o) Shuey would be very similar to A&R. Fig. 29 is another example.
A& R
true curve
Fig. 29
The effect of ray tracing: Calculating accurately the angle of incidences for a given offset
can be a very significant factor in AVO inversion.
It is common in AVO inversion to use “straight ray approximation”. Fig. 30 illustrates the
problem with approximation. The angle of incidence may be quite different from reality when
using simplified approximations.
Fig. 31
In Fig. 31 you can see that for a given offset and a given traveltime, ray tracing with a
horizontal layer model or a dipping layer model can produce very significant differences in the
resulting angle. It is therefore quite important to ray trace using the best approximation to the
real model.
Another factor that could affect the angle calculation is the use of real topography versus
using static correction. It may be important to take topography into account rather than ray
tracing from datum. This effect is illustrated in Fig. 32.
The example in Fig. 33 illustrates the effect of error in angle of incidence calculation on AVO
inversion.
Shuey curve
Fig. 33
This is a synthetic example of an AVO gather. In Fig. 33 ray tracing was performed with the
correct model. You can see a very good fit of the A&R curve.
data
Fig. 34: AVO with wrong background model (error in reflector dip of 10o)
In Fig. 34 ray tracing was performed with a wrong model; there is an error of 10o in the dip of
the reflector. Note that now the A&R curve does not fit the data. Note that the data curve,
which is the same data in both cases, looks different. This is because the mapping of data
from offsets to angle of incidence is different in both cases (Fig. 33 versus Fig. 34). It is wrong
in the second case.
The A&R formula approximates the Zoeppritz equation. The exact Zoeppritz equation which defines
reflection coefficients is given in the following figures. Note that this matrix form defines reflection
coefficients for all wave reflection and transmission combinations.
‘PS’ means down going P-wave reflected S-wave, etc.
Fig. 35
Aki, K., and Richards, P.G., 1980, Quantitative seismology, Theory and methods, Volume 1. W.H.
Freeman and Company.
Bortfeld R., 1961, Approximation to the reflection and transmission coefficients of plane longitudinal and
transverse waves: Geophys. Prosp., 9, 485-502.
Castagna, J., Batzle, M., Eastwood, R., 1985, Relationships between compressional-wave and shear
wave velocities in clastic silicate rocks. Geophysics, 50, 571-581.
Castagna, J.P. Batzle, M.L. and Eastwood, R.L. 1985, Relationships between compressional-wave and
shear-wave velocities in clastic silicate rocks, Geophysics 50, 571-581.
Castagna, Batzle and Kan, 1993. Rock physics: the link between rock properties and AVO response,
Offset Dependent Reflectivity: Theory and Practice of AVO Analysis, Castagna and Backus, pp. 152-154.
Gardner, G.H.F., Gardner, L.W. and Gregory, A.R. 1974, Formation velocity and density - the diagnostic
basics for stratigraphic traps, 1974 Geophysics 39, 770-780.
Gassman, F., 1951, Elastic waves through a packing of spheres. Geophysics, 16, 673-685.
Goodway B., Chen T., Downton J. Improved AVO fluid detection and lithology discrimination using Lame
petrophysical parameters; λρ , µρ , and λ ⁄ µ fluid stack, from P and S inversions. CSEG national
convention, 1997, Expanded Abstracts, p. 183-186.
Koefoed, O., 1995, On the effect of Poisson’s ratios of rock strata on the reflection coefficients of plane
waves. Geophys. Prosp., 3, 381-387.
Krail, P.M., and Brysk, H., 1983, Reflection of spherical seismic waves in elastic layered media:
Geophysics, 48, 655-664.
Movco, G., Mukerji, T. and Dvorkin, J. 1998, The Rock Physics Handbook, Cambridge University Press.
Ostrander, W.J., 1984, Plane wave reflection coefficients for gas sands at nonnormal angles of incidence,
Geophysics 49, 1637-1648.
Shuey, R.T., 1985, A simplification of the Zoeppritz equations, Geophysics 50, 609-614.
Smith, G.C. and Gidlow, P.M., 1987, Weighted stacking for rock property estimation and detection of gas,
Geophysical Prospecting 35, 993-1014.
Verm, R. and Hillerman, F., 1995, Lithology color-coded seismic sections: The calibration of AVO
crossplotting to rock properties, The Leading Edge 14, 847-853.
Zoeppritz, K., 1919, Erdbebenwellen VII B, On the reflection and penetration of seismic waves through
unstable layers: Gottinger Nachr., 1,66-84.