You are on page 1of 332

AVO Workshop Part 1

Bali, Indonesia
January 23, 2008

Table of Contents - Part 1


Overview of the AVO process
Theory : Rock Physics & Fluid Replacement Modeling
Exercise : The Colony Gas Sand Reading in the Logs
Theory : AVO Theory & Modeling
Exercise : The Colony Gas Sand Creating Synthetics
Theory : AVO Analysis on Seismic Data
Exercise : The Colony Gas Sand Calculating AVO Attributes
Theory : Cross Plotting AVO Attributes
Exercise : The Colony Gas Sand Cross Plotting AVO Attributes
Theory : AVO Case Study Onshore Texas Example

1/2/2008

Table of Contents Part 2


Theory : AVO Inversion - Elastic Impedance
Theory : AVO Inversion - Independent AVO Inversion and Lambda-Mu-Rho
Theory : AVO Inversion - Simultaneous Inversion
Exercise : The Colony Gas Sand Simultaneous Inversion
Theory : AVO Case Study Simultaneous Inversion on the Marlin Field
Theory : Processing Issues in AVO
Exercise : Gulf Coast Exercise 1 Data Preparation
Exercise : Gulf Coast Exercise 2 AVO Modeling
Exercise : Gulf Coast Exercise 3 AVO Analysis
Theory : AFI: Analyzing uncertainty in AVO
Theory: Summary
References

1/2/2008

Overview of the AVO Process

This tutorial is a brief introduction to the Amplitude Variations with


Offset, or Amplitude Versus Offset (AVO) method.

We will briefly review how the interpretation of seismic data has


changed through the years.

We will then look at why AVO was an important step forward for the
interpretation of hydrocarbon anomalies.

Finally, we will show why the AVO response is closely linked to the
rock physics of the reservoir.

1/2/2008

A Seismic Section

The figure above shows a stacked seismic section recorded over the shallow
Cretaceous in Alberta. How would you interpret this section?
1/2/2008

Structural Interpretation

Your eye may first go to an interesting seismic event between 630 and 640 ms. Here, it
has been picked and called H1. A seismic interpreter prior to 1970 would have looked
only at structure and perhaps have located a well at CDP 330.
1/2/2008

Gas Well Location

And, in this case, he or she would have been right! A successful gas well was drilled
at that location. The figure above shows the sonic log, integrated to time, spliced on
the section. The gas sand top and base are shown as black lines on the log.
1/2/2008

Bright Spots

But this would have been a lucky guess, since structure alone does not tell you that a
gas sand is present. A geophysicist in the 1970s would have based the well on the
fact that there is a bright spot visible on the seismic section, as indicated above.
1/2/2008

What is a Bright Spot?


Surface

Geology

Seismic

Seismic
raypath

Shale
Gas Sand
Interface at
depth = d

1 V1

R0 =

2 V2

Reflection at time
t = 2d/V1

2V2 1V1
2V2 + 1V1

Seismic
Wavelet

To understand bright spots, recall the definition of the zero-offset reflection


coefficient, shown in the figure above. R0 , the reflection coefficient, is the amplitude
of the seismic peak shown. Note also that the product of density, , and P-wave
velocity, V, is called acoustic impedance.
1/2/2008

The figure on the right,


taken from Gardner et al.
(Formation velocity and
density - The diagnostic
basics for stratigraphic
traps: Geophysics, 1974)
shows that there is a big
difference between shale
and gas sand velocity at
shallow depths. The paper
also shows that density and
velocity are approximately
related by the equation

Difference between shale and gas sand


velocity at shallow depth.

= 0.23 V 0.25
Thus, we would expect a big
reflection coefficient, or
bright spot, for shallow
gas sands.
1/2/2008

10

The AVO Method

1/2/2008

Unfortunately, bright
spots can be caused
by lithologic
variations as well as
gas sands. This lead
geophysicists in the
1980s to start
looking at pre-stack
seismic data. The
amplitude increase
with offset shown
here was predicted
by Ostrander (Planewave reflection
coefficients for gas
sands at nonnormal
angles-of-incidence:
Geophysics, 1984) for
certain gas sands
(Class 3, as we will
11
discuss later).

What causes the AVO Effect?


Surface

2 1

1 VP1 VS1

Reflector

2 VP2 VS2
As shown above, the traces in a seismic gather reflect from the subsurface at
increasing angles of incidence . The first order approximation to the reflection
coefficients as a function of angle is given by adding a second term to the zero-offset
reflection coefficient:

R( ) = R0 + B sin2

B is a gradient term which produces the AVO effect. It is dependent on changes in


density, , P-wave velocity, VP, and S-wave velocity, VS.
1/2/2008

12

Why is S-wave Velocity Important?


As just shown, the gradient
term is dependent on
density, P and S-wave
velocity. The reason that Swave velocity has such an
impact is shown on the left,
where P and S-wave velocity
are shown as a function of
gas saturation in the
reservoir. Note that P-wave
velocity drops dramatically,
but S-wave velocity only
increases slightly (why?).
This will be discussed
thoroughly in the next
chapter.

1/2/2008

13

AVO Modeling
P-wave

Density

S-wave

Poissons
ratio

Synthetic

Offset Stack

Based on AVO theory and the rock physics of the reservoir, we can perform AVO
modeling, as shown above. Note that the model result is a fairly good match to the
offset stack. Also note that Poissons ratio is a function of Vp/Vs ratio and will be
discussed in the next chapter.
1/2/2008

14

AVO Attributes
Intercept: A

Gradient: B
AVO Attributes are
used to analyze
large volumes of
seismic data,
looking for
hydrocarbon
anomalies.
1/2/2008

15

Cross-Plotting of Attributes
Gradient (B)

One of the AVO methods that we will be


discussing later in the course involves crossplotting the zero-offset reflection coefficient (A),
versus the gradient (B), as shown on the left.
As seen in the figure below, the highlighted zones
correspond to the top of gas sand (pink), base of
gas sand (yellow), and a hard streak below the
gas sand (blue).

Intercept (A)

1/2/2008

16

AVO Inversion
A very important new
tool combines
Inversion with AVO
Analysis to enhance
the reservoir
discrimination.

Far Inversion

Near Inversion

1/2/2008

17

Summary of AVO Methodology


Input NMO-corrected Gathers

Modeling
Wave
Equation

Recon
Methods

Zoeppritz
Partial
Stacks

Intercept
Gradient
Attributes

Inversion
Elastic
Impedance

LMR

Simultaneous
Inversion

Cross
Plots

1/2/2008

18

Conclusions

Seismic interpretation has evolved over the years, from strictly structural
interpretation, through bright spot identification, to direct hydrocarbon
detection using AVO.

In this course we will elaborate on the ideas that have been presented in
this short introduction.

As a starting point, the next chapter will discuss the principles of rock
physics in more detail.

We will then move to AVO modeling.

Finally, we will look at AVO analysis on real seismic data.

In each case, we will first look at the theory and then perform a
workstation example.

1/2/2008

19

Rock Physics & Fluid


Replacement Modeling

Basic Rock Physics


The AVO response is dependent on the properties of P-wave velocity (VP),
S-wave velocity (VS), and density () in a porous reservoir rock. As shown
below, this involves the matrix material, the porosity, and the fluids filling
the pores:

Rock Matrix
1/2/2008

Pores / Fluid
21

Density
Density effects can be modeled with the following equation:

sat = m( 1 ) + w S w + hc( 1 S w )
where : = density,

= porosity,
S w = water saturation,
sat,m,hc, w = saturated, matrix,
hydrocarbon, water subscripts.
This is illustrated in the next graph.

1/2/2008

22

Density versus Water Saturation


Here is a plot of density
vs water saturation for a
porous sand with the
parameters shown,
where we have filled the
pores with either oil or
gas.

1/2/2008

Sandstone with Porosity = 33%


Densities (g/cc): Matrix = 2.65, Water = 1.0,
Oil = 0.8, Gas = 0.001
2.2
2.1
2
Density

In the section on AVO we


will model both the wet
sand and the 50%
saturated gas sand.
Note that these density
values can be read off
the plot and are:
wet = 2.11 g/cc
gas = 1.95 g/cc

Density vs Water Saturation

1.9
1.8
1.7
1.6
0
Oil

0.1

0.2
Gas

0.3

0.4

0.5

0.6

0.7

0.8

0.9

Water Saturation
23

P and S-Wave Velocities


Unlike density, seismic velocity involves the deformation of a rock as a
function of time. As shown below, a cube of rock can be compressed, which
changes its volume and shape or sheared, which changes its shape but not
its volume.

1/2/2008

24

P and S-Wave Velocities


This lead to two different types of velocities:
P-wave, or compressional wave velocity, in which the direction of
particle motion is in the same direction as the wave movement.
S-wave, or shear wave velocity, in which the direction of particle
motion is at right angles to the wave movement.

P-waves

1/2/2008

S-waves

25

Velocity Equations using and

The simplest forms of the P and S-wave velocities are derived for
non-porous, isotropic rocks. Here are the equations for velocity
written using the Lam coefficients:

+ 2
VP =

VS =

where: = the first Lam constant,


= the second Lam constant,
and
= density.

1/2/2008

26

Velocity Equations using K and


Another common way of writing the velocity equations is with
bulk and shear modulus:

VP =

4
K+
3

VS =

where: K = the bulk modulus, or the reciprocal of compressibility.


= the shear modulus, or the second Lam constant,
and
= density.

1/2/2008

27

Poissons Ratio
A common way of looking at the ratio of VP to VS is to use Poissons ratio,
defined as:

2
=
2 2
VP
where : =
VS

The inverse to the above formula, allowing you to derive VP or VS from


, is given by:

2 2
=
2 1
1/2/2008

28

Poissons Ratio vs VP/VS ratio


Vp/Vs vs Poisson's Ratio
0.5

Poisson's Ratio

0.4
0.3
0.2
0.1
0
-0.1
-0.2
0
Gas Case
1/2/2008

Wet Case

10

Vp/Vs
29

Poissons Ratio
From the previous slide, note that there are several values of
Poissons ratio and VP/VS ratio that are important to remember.
If VP/VS = 2, then = 0
If VP/VS = 1.5, then = 0.1 (Gas Case)
If VP/VS = 2, then = 1/3 (Wet Case)
If VP/VS = , then = 0.5 (VS = 0)

1/2/2008

30

Velocity in Porous Rocks


Velocity effects can be modeled by the volume average equation as
seen below and in the next figure:

t sat = t m( 1 ) + t w S w + t hc( 1 S w )
where t = 1 / V
Unfortunately, the above equation does not hold for gas sands, and this
lead to the development of other equations.

1/2/2008

31

Velocity vs Saturation using Volume


Averaging
Velocity vs Water Saturation
Wyllie's Equation

Unfortunately, this
equation does not hold
for gas sands, and this
lead to the development
of the Biot-Gassmann
equations.

Porosity = 33%
Vmatrix = 5700 m/s, Vw = 1600 m/s,
Voil = 1300 m/s, Vgas = 300 m/s.
3500
3000
Velocity (m/sec)

A plot of velocity vs water


saturation using the
previous equation for a
porous sand with the
parameters shown, where
we have filled the pores
with either oil or gas.

2500
2000
1500
1000
500
0
Oil

1/2/2008

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9


Gas

Water Saturation

32

The Biot-Gassmann Equations


Unfortunately, the volume average equation gives incorrect results for gas
sands. Independently, Gassmann (1951) and Biot (1956), developed a more
complete theory of wave propagation in fluid saturated rocks, by deriving
expressions for the saturated bulk and shear moduli, and substituting into
the regular equations for P and S-wave velocity:

VP _ sat =

K sat

4
+ sat
3

sat

VS _ sat

sat
=
sat

Note that sat is found using the volume average equation discussed
earlier.

1/2/2008

33

The Biot-Gassmann Equations


To understand the Biot-Gassmann equations, let us update the figure we saw
earlier to include the concepts of the saturated rock (which includes the insitu fluid) and the dry rock (in which the fluid has been drained.)

Dry rock
frame, or
skeleton
(pores
empty)
Rock Matrix
1/2/2008

Saturated
Rock
(pores full)

Pores and fluid


34

Biot-Gassmann - Shear
Modulus
In the Hampson-Russell AVO program, Biot-Gassmann analysis is done
using the FRM (Fluid Replacement Modeling) option. Let us first look at
some theory and then consider several practical considerations when using
the FRM option.
In the Biot-Gassmann equations, the shear modulus does not change for
varying saturation at constant porosity. In equations:

sat = dry
where sat = the shear modulus of the saturated rock,
and

1/2/2008

dry = the shear modulus of the dry rock.

35

Biot-Gassmann Saturated Bulk


Modulus
The Biot-Gassmann bulk modulus equation is as follows:
2

(1)

K sat = K dry

K dry
1

Km

+
1 K dry
+
2
K fl
Km
Km

Mavko et al, in The Rock Physics Handbook, re-arranged the above


equation to give a more intuitive form:

(2)

K dry
K fl
K sat
=
+
K m K sat K m K dry ( K m K fl )

where sat = saturated rock, dry = dry frame, m = rock matrix, fl = fluid,
and = porosity.
1/2/2008

36

Biots Formulation
Biot defines the parameters (the Biot coefficient) and M (the fluid modulus),
given by:

= 1

K dry
Km

, and


1
=
+
,
M K fl
Km

Equation (1) on the previous slide then can be written as:

If = 0 (or Kdry = Km) this equation simplifies to:


If = 1 (or Kdry= 0), this equation simplifies to:

K sat = K dry + 2 M

K sat = K dry
1
1
=
+
K sat K fl
Km

Physically, = 0 implies we have a non-porous rock, and = 1 implies we have


particles in suspension (and the formula given is called Woods formula). These are
the two end members of a porous rock.
1/2/2008

37

The Rock Matrix Bulk Modulus


We will now look at how to get estimates of the various bulk modulus terms
in the Biot-Gassmann equations, starting with the bulk modulus of the solid
rock matrix. Values will be given in gigaPascals (GPa), which are
equivalent to 1010 dynes/cm2.

The bulk modulus of the solid rock matrix, Km is usually taken from
published data that involved measurements on drill core samples. Typical
values are:

Ksandstone = 40 GPa,
Klimestone = 60 GPa.
1/2/2008

38

The Fluid Bulk Modulus


The fluid bulk modulus can be modeled using the following equation:

Sw 1 Sw
1
=
+
K fl K w
K hc
where K fl = the bulk modulus of the fluid,
K w = the bulk modulus of the water,
and

K hc = the bulk modulus of the hydrocarbon.

Equations for estimating the values of brine, gas, and oil bulk modulii are
given in Batzle and Wang, 1992, Seismic Properties of Pore Fluids,
Geophysics, 57, 1396-1408. Typical values are:

Kgas = 0.021 GPa, Koil = 0.79 GPa, Kw = 2.38 GPa


1/2/2008

39

Estimating Kdry
The key step in FRM is calculating a value of Kdry. This can be done in
several ways:
(1) For known VS and VP, Kdry can be calculated by first calculating Ksat
and then using Mavkos equation, given earlier.
(2) For known VP, but unknown VS, Kdry can be estimated by:
(a) Assuming a known dry rock Poissons ratio dry. Equation (1) can
then be rewritten as a quadratic equation in which we solve for Kdry.
(b) Using the mudrock equation to estimate the wet case and then
using a procedure developed by Mavko et al. (Fluid Substitution:
Estimating changes in VP without knowing VS, Geophysics, Nov-Dec,
1995) to calculate the hydrocarbon case. (See Appendix)

1/2/2008

40

Data Examples

In the next few slides, we will look at the computed responses for
both a gas-saturated sand and an oil-saturated sand using the BiotGassmann equation.
We will look at the effect of saturation on both velocity (VP and VS)
and Poissons Ratio.
Keep in mind that this model assumes that the gas is uniformly
distributed in the fluid. Patchy saturation provides a different
function. (See Mavko et al: The Rock Physics Handbook.)

1/2/2008

41

Velocity vs Saturation of Gas


A plot of velocity vs water
saturation for a porous gas
sand using the Biot-Gassmann
equations with the parameters
shown.

1/2/2008

2600
2400
2200
Velocity (m/s)

In the section on AVO we will


model both the wet sand and
the 50% saturated gas sand.
Note that the velocity values
can be read off the plot and
are:
VPwet = 2500 m/s
VPgas = 2000 m/s
VSwet = 1250 m/s
VSgas = 1305 m/s

Velocity vs Water Saturation - Gas Case


Sandstone with Phi = 33%, Density as previous figure for gas,
Kmatrix = 40 Gpa, Kdry = 3.25 GPa, Kw = 2.38 Gpa,
Kgas = 0.021 Gpa, Shear Modulus = 3.3. Gpa.

2000
1800
1600
1400
1200
1000
0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

Sw
Vp

Vs

42

Poissons Ratio vs Saturation of Gas


Poisson's Ratio vs Water Saturation - Gas Case
Sandstone with Phi = 33%, Density as previous figure for gas,
Kmatrix = 40 Gpa, Kdry = 3.25 GPa, Kw = 2.38 Gpa,
Kgas = 0.021 Gpa, Shear Modulus = 3.3. Gpa.

A plot of Poissons ratio vs


water saturation for a porous
gas sand using the BiotGassmann equations with the
parameters shown.

0.5

In the section on AVO we will


model both the wet sand and
the 50% saturated gas sand.
Note that the Poissons ratio
values can be read off the plot
and are:
wet = 0.33
gas = 0.12

Poisson's Ratio

0.4

0.3

0.2

0.1

0
0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

Sw

1/2/2008

43

Velocity vs Saturation of Oil


Velocity vs Water Saturation - Oil Case

A plot of velocity vs water


saturation for a porous oil
sand using the BiotGassmann equations with
the parameters shown.

2600
2400
2200
Velocity (m/s)

Note that there is not much


of a velocity change.
However, this is for dead
oil, with no dissolved gas
bubbles, and most oil
reservoirs have some
percentage of dissolved
gas.

Sandstone with Phi = 33%, Density as previous figure for oil,


Kmatrix = 40 Gpa, Kdry = 3.25 GPa, Kw = 2.38 Gpa,
Koil = 1.0 Gpa, Shear Modulus = 3.3. Gpa.

2000
1800
1600
1400
1200
1000
0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

Sw
Vp
1/2/2008

Vs
44

Poissons Ratio vs Saturation of Oil


Poisson's Ratio vs Water Saturation - Oil Case
Sandstone with Phi = 33%, Density as previous figure for oil,
Kmatrix = 40 Gpa, Kdry = 3.25 GPa, Kw = 2.38 Gpa,
Koil = 1.0 Gpa, Shear Modulus = 3.3. Gpa.

A plot of Poissons ratio vs


water saturation for a porous
oil sand using the BiotGassmann equations with the
parameters shown.

0.4

Poisson's Ratio

Note that there is not much of


a Poissons ratio change.
However, again this is for
dead oil, with no dissolved
gas bubbles, and most oil
reservoirs have some
percentage of dissolved gas.

0.5

0.3

0.2

0.1

0
0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

Sw

1/2/2008

45

Tips for Using Gassmanns Equation


CAUTIONS:
Rocks with large Km and Kdry values (most carbonates) appear insensitive
to saturation changes in Gassmann theory.
Gassmann assumed that pore pressure remains constant during wave
propagation. This implies fluids are mobile between pores and all stress is
carried by Kdry.
This assumption is violated at high frequencies in highly variable and
compressible pore systems.
Carbonates with an abundance of crack-type pores and heterogeneous
pore systems are not suitable for standard Gassmann theory.

1/2/2008

46

Patchy Saturation
When multiple pore fluids are present, Kfl is usually calculated by a Reuss
averaging technique:

1
S w So S g
=
+
+
K fl K w Ko K g

This averaging
technique assumes
uniform fluid
distribution!
-Gas and liquid must
be evenly distributed
in every pore.
1/2/2008

Bulk modulus (Gpa)

This method heavily biases compressibility of the combined fluid to


the most compressible phase.
Kfl vs Sw and Sg

3
2.5
2
1.5
1
0.5
0
0

0.25

0.5

0.75

Water saturation (fraction)

47

Patchy Saturation
When fluids are not uniformly mixed, effective modulus values cannot be
estimated from Reuss averaging. Uniform averaging of fluids does not
apply.
When patch sizes are large with respect to the seismic wavelength, Voigt
averaging gives the best estimate of Kfl (Domenico, 1976).

K fl = S w K w + So K o + S g K g
When patch sizes are of intermediate size, Gassmann substitution should be
performed for each patch area and a volume average should be made
(Dvorkin et al, 1999). This can be approximated by using a power-law
averaging technique.

1/2/2008

48

Patchy Saturation
Gassmann predicted velocities
Unconsolidated sand matrix
Porosity = 30%
100% Gas to 100% Brine saturation

Vp (km/s)

2.5
2.3
Patchy
Voigt
Reuss

2.1
1.9
1.7
1.5
0

1/2/2008

0.25

0.5

0.75

Water Saturation (fraction)

1
49

The Mudrock Line

The mudrock line is a linear relationship between VP and VS derived


by Castagna et al (1985):

VP = 1.16 VS + 1360 m / s
Note that for a constant Poissons ratio, the intercept is zero:

VP =

2 2
VS
2 1

This will be illustrated in the next few slides.

1/2/2008

50

The Mudrock Line

ARCOs original mudrock derivation


(Castagna et al, Geophysics, 1985)
1/2/2008

51

The Mudrock Line


6000
5000
Mudrock Line

4000
3000

Gas Sand

VP (m/s)
2000
1000

0
0
1/2/2008

1000

VS(m/s)

2000

3000

4000
52

The Mudrock Line


6000
5000

= 1/3

Mudrock Line

or
VP/VS = 2

4000
3000

Gas Sand

VP (m/s)
2000
1000

0
0
1/2/2008

1000

VS(m/s)

2000

3000

4000
53

The Mudrock Line


6000
5000

= 1/3 or

Mudrock Line

VP/VS = 2

4000
3000

Gas Sand

VP (m/s)
2000

= 0.1 or
VP/VS = 1.5

1000

0
0
1/2/2008

1000

VS(m/s)

2000

3000

4000
54

Using the Biot-Gassmann Equations


(1) Fluid Substitution
The basic use of the Biot-Gassmann equations is to substitute or replace the fluids
in a set of target layers with another set of fluids.
In this case, VP, VS, and must all be known for the input logs, along with the fluid
content (SW). Generally all three logs are changed within the target zone.

Output Logs

Input Logs

VP
1/2/2008

SW = 50%

VS

VP

SW = 100%

VS
55

Using the Biot-Gassmann Equations


(2) Calculating Vs
The second use of the Biot-Gassmann equations is to calculate a VS curve, which
has not been measured in the well. Either KDRY is assumed known or the mudrock
equation is assumed to hold for wet sands.
In this case, VP and must both be known, along with the fluid content (SW). The VP
and logs are unchanged, and a new VS log is created.

Output Logs

Input Logs

VP
1/2/2008

SW = 50%

VP

SW = 50%

VS
56

Using the Biot-Gassmann Equations


In fact, the calculation of Vs requires 2 separate equations: Castagnas equation for
the wet sand/shales (outside the target). Biot-Gassmann inside the target. The
calculation is performed in 2 steps:
(1)

Use Castagnas
equation first to create
a VS log, which is
accurate everywhere
except in the target.

VP

(2)

VP

VS-Cast

Use Biot-Gassmann to
calculate the correct
VS values inside the
target.

VP
1/2/2008

VS-Cast

VP

VS
57

Conclusions
An understanding of rock physics is crucial for the interpretation of AVO
anomalies.
The volume average equation can be used to model density in a water
sand, but this equation does not match observations for velocities in a gas
sand.
The Biot-Gassmann equations match observations well for unconsolidated
gas sands.
When dealing with more complex porous media with patchy saturation, or
fracture type porosity (e.g. carbonates), the Biot-Gassmann equations do
not hold.
The ARCO mudrock line is a good empirical tool for the wet sands and
shales.
1/2/2008

58

Appendix: Calculation of VS using


Castagna Assumption
1) Calculate density for 100% brine
saturation:

wet = br + m ( 1 - )
2) Calculate input P wave modulus:

5) Calculate Vp_wet

Vp wet =

M wet

wet

6) Calculate Vs_wet from Vp_wet

M = Vp2

Vswet =Ac Vpwet + Bc

3) Calculate matrix P wave


modulus:

4
Mm = Km + m
3
4) Adjust P wave modulus to 100%
water:

d=

wet
Vs = Vs wet *

M
K fl
K br

+
M m M * ( M m K fl ) * ( M m K br )

Mwet = d *
1/2/2008

7) Calculate Vs_input from Vs_wet

Mm
1+ d

59

Appendix: Calculation of VS using


Castagna Assumption
8) Calculate K and m from input data:

10) Calculate Ksat with new fluid:

= Vs * ;

Kdryout
Kflout
a=
+ out
out
Km Kdry
*(Km Kflout )

4
K = *Vp *
3
2

9) Obtain K_dry:

a =

K
K fl

Km K
* ( K m K fl )

a
Kdry = Km *
1+ a

a
K = Km*
1+ a
out

11) Get new density:

out = fl out * out + m *(1 out )


12) Finally the new velocities!
Vp out =

1/2/2008

K out +

4 out

3
;
out

Vs out

out
=
out
60

Appendix:
Averaging Multiple Minerals
We may also want to average multiple mineral components to produce a
composite material. The simplest way to do this is to use a Voigt or
Reuss average. If we let f1 be the fraction of mineral 1, f2 be the fraction
of mineral 2 (where f1 + f2 = 1), M1 be the modulus of mineral 1 (bulk or
shear) and M2 be the modulus of mineral 2, then the Voigt average is
given by:

M V = f1 M 1 + f 2 M 2
The Reuss average is given by:

1
M 1M 2
f1
f2
=
+
MR =
f1 M 2 + f 2 M 1
M R M1 M 2
1/2/2008

61

Appendix:
Hashin-Shtrikman Bounds
The Voigt and Reuss bounds give extreme values. Another approach is to
use Hashin-Shtrikman bounds, which are different for the bulk and shear
modulus components. If mineral 1 is stiffer than mineral 2, then the upper
bound is given by (Mavko et al.):

K HS

f2
= K1 +
( K 2 K1 ) 1 + f1 ( K1 + ( 4 / 3) 1 ) 1

HS = 1 +

f2
2 f1 ( K1 + 2 1 )
1
( 2 1 ) +
51 ( K1 + ( 4 / 3) 1 )

The lower bounds are given by reversing the order of the two minerals in the
equations given above. An example is shown on the next page.
1/2/2008

62

Appendix
Comparing the Bounds

The figures above show the effect of Voigt, Reuss and Hashin-Shtrikman
upper and lower bounds for materials with K1 = 60 GPa, K2 = 40 GPa, 1 = 45
GPa, and 2 = 15 GPa. Note that the H-S bounds are between the Voigt and
Reuss bounds. In the software, we use the average of the H-S bounds.
1/2/2008

63

The Colony Gas Sand


Reading in the Logs

Our first set of exercises comes from the Colony sand formation, a Cretaceous sand
from Western Canada.
The target is a thin, 8 meter
thick, gas sand.

P-wave

Density

S-wave

Poissons
Ratio

We have a single well with


measured sonic and
density logs.
The sand is known to
contain 50% water, and 50%
gas.
In this exercise, we will
create the S-wave log,
using Castagnas equation
and Biot-Gassmann
analysis.
1/2/2008

65

Start the GEOVIEW program by typing


geoview in the command window (Unix) or
selecting Geoview from Start / Programs /
HRS applications (Windows).
We are going to create a new database, so
click on New when you see the Opened
Database List window, as shown. Then
click on OK.

On the next menu, name the database avo_class, and click on OK.

1/2/2008

66

A database is a collection of wells. Within each well there may be any number of
logs. These logs may be of any type. They need not all be sampled at the same
depth values, and they need not be sampled uniformly. The GEOVIEW Well Explorer
window, which allows us to view the contents of a database, currently is empty:

1/2/2008

67

As well as the Well Explorer window, you will also see the program launcher bar, as
shown below:

The program launcher bar is used to launch any of the suite of Hampson-Russell
software programs that your company owns, as well as to re-launch the Well
Explorer window if this window gets closed.
Later in this exercise, we will launch the AVO program. But first we will read a well
into the database.

1/2/2008

68

Reading in Well Data in LAS Format


In the next set of exercises, we will read in a real set of well logs and a pre-stack
seismic line. We will then create a synthetic CDP gather and compare it with the real
data.
To read the logs into the GEOVIEW Well Explorer, click on Import Data / Logs, Check
Shots, Tops, Deviated Geometry from Files:

1/2/2008

69

On the File Import menu that appears, select the directory containing the AVO
workshop data. (You will have to check with the instructor to find the right
directory, as this varies from class to class). Choose the avo_well.las file:

Make sure that the the Log File Format is set to LAS, then click on Next >>.

1/2/2008

70

By default, the program will create a new well called AVO_WELL, which is the well
name stored in the file headers. We will accept that default. Click on Next >>.

The next page allows you to specify parameters for this well. Click on Next >> to
accept the defaults:

1/2/2008

71

The final page lists all the curves in the LAS file and allows you to overwrite
parameters. Click on Ok to accept all the defaults:

Note that the program will tell you that the logs have
been successfully imported with the message shown
on the right. Click Ok on this message.

Finally, a menu appears confirming which units you


wish to use for the default display settings. Click Ok
to accept the default, metric, for this well.

1/2/2008

72

Now that all the logs have been inserted into the database, there are a number of
ways of examining them. First, click on the field containing the name AVO_WELL in
the GEOVIEW Well Explorer window and then click on the Display Well button, as
shown below:

1/2/2008

73

This will cause a


display window to
appear, showing
you all the logs and
tops in that well:

1/2/2008

74

If you click on View / Display Options at the top of the Log Display Window, you will
see a menu which allows you to change the look of this display:

This is an example of
a notebook menu,
which allows you to
change pages by
clicking on the tabs at
the top. Notice that
the available options
are Layout, Scale &
Details, and
Synthetics.

1/2/2008

75

The default Layout


tab in this menu
allows you to overlay
logs. For example,
select the Computed
Impedance and
Computed
Reflectivity logs to
overlay the P-wave
log in the same track.
Click on Apply.

1/2/2008

76

The three logs are


now displayed in a
single track at
three different
scales and in three
different colors.

1/2/2008

77

Now click on the Curves tab and set the amplitude range for the Density log from 1.8 to
2.7 and for the SP log from -150 to -50 as shown below:

After you have


changed these
numbers click
on Save
settings as
project
template. This
will mean that
all future
displays will
use these log
ranges.

1/2/2008

78

Click on Ok on the
Parameter menu, and
the Log Display
window is redrawn
as shown:

1/2/2008

79

The GEOVIEW Well Explorer provides another way of looking at the contents of the
database:

The Table View contains one line for each well in the database. Note that any of the
parameters on this window may be changed.
1/2/2008

80

Now click on the arrow to the left of the


field showing the name AVO_WELL:

Now you see a list of all the logs in this well. Once again, any of these parameters
can be changed.
1/2/2008

81

Applying a Check Shot Correction


Now we will apply a check shot correction to AVO_WELL. First, we must enter check
shot values as a log. From the Table View of the Well Explorer window, click on Log
Options / Create a new log in table / Check Shot:

1/2/2008

82

A new row will appear at the bottom of the Table View, showing the new Log Name,
the Log Type and the Amplitude Units:

To begin entering values, click on the arrow to the left of the Log Name field:

1/2/2008

83

Fill in the table as shown:

This check shot contains two pairs of values consisting of a depth and a two-way
time in milliseconds. Click on Update on this menu to add the check shot to the
database, then click on Yes on the subsequent dialog to confirm the addition of this
log, as shown below:

1/2/2008

84

Even though the check


shot values have been
added, the check shot will
not be applied to the sonic
log until we perform that
operation. Note that the
check shot log has now
been added to the log
display

1/2/2008

85

To apply the check shot correction, click on


Option / Check Shot Correction:
The Check Shot Parameters menu and the Check Shot Analysis window appear:

1/2/2008

86

There are a number of modifications which could be made to the automatic


calculation. One modification is to delete points from the check shot data and
reapply the correction. Another modification is to use a polynomial interpolation
between the points to prevent artificial discontinuities in the corrected log. For now,
we will accept the default parameters. Click on Ok on the Check Shot Analysis
window. This menu now appears:

Click on Ok on the menu. This will create a new check shot corrected sonic log, with
the name P-wave_chk.

1/2/2008

87

Launching the AVO Program


Now we will begin creating an AVO model from
the logs we have just entered. Click on the
AVO/AVO button from the GEOVIEW main
window to launch AVO.

Choose the option to Start New Project:

Call the new project colony:

1/2/2008

88

This will initiate a blank AVO program window, as shown below.

1/2/2008

89

Modeling with the AVO Program


When the AVO window appears, click on Modeling
/ Single Well:

Choose the well (AVO_WELL)


that we have just imported to
do the modeling and click on
Open:

1/2/2008

90

On this menu, we wish to use the default (i.e., active) logs as input:
Click OK.

The following information window


appears, telling you that a shear
wave log will be calculated using
transforms. Click OK.

1/2/2008

91

Creating the S-Wave Log


On the next series of menus, we will accept the defaults for creating the shear
wave log. We will use Castagnas equation:

Click Next >>

1/2/2008

92

We will apply the equation to the check shot corrected sonic log and call the
output log S-wave_Castagna. Change the name as shown below:

Click Next >>


1/2/2008

93

We will use the default Castagna coefficients:

Finally, click OK.


1/2/2008

94

The AVO Modeling window now shows the created S-wave log and the calculated
Poissons ratio.

1/2/2008

95

Fluid Replacement Modeling (FRM)


The S-wave log calculated so far, has used the Castagna mud-rock equation. This is
only appropriate for wet sands and shales. This means that the S-wave velocity
calculated within the target sand layer has the wrong value.
We will now use the Biot-Gassmann equation to replace the calculated S-wave
velocity value, in that interval, with one more appropriate for the gas case.
For this calculation, we need the further information about the actual fluid content in
the sand. As it happens, the real water saturation is 50%.

To perform Fluid Replacement Modeling,


click on the FRM button:

1/2/2008

96

On the first page, we specify


what is known about the
input logs.
Since the porosity was not
measured, we will calculate
it from the input density log.
We are specifying a 2-phase
Fluid Composition,
consisting of brine and gas.
Finally, change the Water
Saturation to 50%, then
click on Next >> to accept
this value.

1/2/2008

97

The second page sets the


depth zone over which the
analysis will be performed.
The most convenient way is
to specify the top and base
of the gas sand with tops,
as shown. Every depth
sample in the range 633 to
640 m will be modified.
Click on Next >>.

1/2/2008

98

This page specifies the petrophysical parameters. These are the density and moduli
of each of the constituent components: matrix, hydrocarbon, and brine.
Note that we will
assume the other
50% of the fluid is
composed of gas.

The default
parameters are
book values.
One way to
modify them is to
click on Display
Fluid properties
calculator

1/2/2008

99

The Batzle-Wang menu allows


you to calculate fluid properties
from more fundamental
measurements. In this case,
we will accept the default gas
and brine parameters.
Click on Cancel on the BatzleWang menu:

1/2/2008

100

The default matrix


parameters have
also been supplied
from book values.

One way to modify


these parameters is
to click on the
Display Matrix
Properties Calculator
button:

1/2/2008

101

The Minerals menu allows you to specify the precise components of the reservoir
matrix and thus calculate the appropriate values of the density and bulk modulus.
To accept the default sand values, click on Cancel on this menu.

1/2/2008

102

Finally, the last page of the FRM


menu sets the output log
parameters. We have specified
that this input well had a water
saturation of 50% and porosities
calculated from the density log.
We could create output logs with
different values of these
parameters. In our case, we
want to keep the same water
saturation and porosity values.
The effect will be to calculate the
S-wave log (Swave_Castagna_FRM)
appropriate to these parameters.
To see the resulting
calculations, click on Apply.

1/2/2008

103

After you click on Apply, the calculated output values for the first layer are printed
here:

To see the results for the entire window, click on the QC Display button:

1/2/2008

104

This display shows all the logs


that have been calculated
within the analysis window.
Scrolling through the window,
we can verify that the
calculation is reasonable.

Finally, click on Ok to
create the new logs:

1/2/2008

105

You are first asked if you are sure


you want to create these logs.
Click on Yes.

The program next asks if you wish


to copy the depth-time curve from
the original log. The answer is Yes,
because we have done check shot
correction and we do not wish to
modify the depth-time relationship,
even if the P-wave log had
changed.

1/2/2008

106

When the new window is plotted, notice the modification in both the S-wave velocity
and Poissons Ratio logs.

1/2/2008

107

Saving the Project


Now click on Project / Save to save the
work done so far:

You will see a menu appear, which asks


whether you wish to save the new created
logs within the GEOVIEW database, or just
within the current project. Click on Yes to
see the list of new logs.

Now a further menu


appears, listing all the new
logs, which have been
created. Click on Ok to
save them all to GEOVIEW.
1/2/2008

108

AVO Theory & Modeling

P and S-Waves

(a)

(b)

(c)

The above diagram shows a schematic diagram of (a) P, or compressional,


waves, (b) SH, or horizontal shear-waves, and (c) SV, or vertical shear-waves,
where the S-waves have been generated using a shear wave source. (Ensley,
1984)
1/2/2008

110

From P and S-Waves to AVO


In the previous slide, the P and SH-waves were generated at the surface
by P and S-wave sources. We could use the differences between the
recorded P and S reflections to discriminate gas-filled sands from wet
sands, using the properties discussed in the last section.
Unfortunately, most seismic surveys record P-wave data only, and Swave data is not available.
However, as shown in the next slide if we record P-wave data at an angle
greater than zero, we produce mode conversion from P to S-wave data.
The objective of this course will be to utilize mode conversion both
quantitatively and qualitatively.

1/2/2008

111

Mode Conversion of an Incident


P-Wave
If > 0, an incident P-wave will produce both P and SV reflected and
transmitted waves. This is called mode conversion.

Incident
P-wave

Reflected
SV-wave = RS(1)

1
1

VP1 , VS1 , 1
VP2 , VS2 , 2

2
2

1/2/2008

Reflected
P-wave = RP(1)

Transmitted
P-wave = TP(1)
Transmitted
SV-wave = TS(1)

112

Utilizing Mode Conversion


But how do we utilize mode conversion? There are actually two ways:
(1) Record the converted S-waves using three-component receivers (in the
X, Y and Z directions). Note that when we analyze the converted waves,
we need to be very careful in their processing and interpretation.
(2) Interpret the amplitudes of the P-waves as a function of offset, or angle,
which contain implied information about the S-waves. This is called the
AVO (Amplitude versus Offset) method.
In the AVO method, we can make use of the Zoeppritz equations, or some
approximation to these equations, to extract S-wave type information from P-wave
reflections at different offsets. Before discussing these equations, the next figures
shows a typical set of gathers over a gas sand and intuitively explain the
relationship between offset and angle.
1/2/2008

113

A Data Example
The box in the top
figure shows the
stacked traces
corresponding to the
five P-wave gathers
shown in the bottom
figure.
Notice that the
amplitudes in the
gathers over the
highlighted region
show an increase in
amplitude as a
function of offset.
This is called an AVO
(Amplitude Variation
with Offset) effect.
1/2/2008

114

Angle and Offset


Offset

Source

Receiver
Surface

Angles

2 1

1 VP1 VS1
2 VP2 VS2

Reflector
Common mid-point

Seismic data is recorded as a function of offset. However, as shown above,


there is a direct relationship between angle and offset, which depends on
velocity. Later, we will show the equations used to do this conversion.
1/2/2008

115

The Zoeppritz Equations


To explain the amplitude change we saw in the mode conversion slide and
on the seismic gathers, Zoeppritz derived the amplitudes of the reflected
and transmitted waves using the conservation of stress and displacement
across the layer boundary, which gives four equations with four
unknowns. Inverting the matrix form of the Zoeppritz equations gives us
the exact amplitudes as a function of angle:

sin 1
RP(1) cos
1
R ( )
S 1 = sin 2
1
TP(1)


TS (1) cos 21

1/2/2008

cos 1

sin 2

cos 2

sin 1
sin 2
cos 2

2
VP1
2VS2VP1
2VS2VP1

cos 21
cos
2

cos
2

1
2

VS1
1VS12
1VS12VP2

VS1
2VP2
2VS2
sin 21
cos 22
sin 22
VP1
1VP1
1VP1

sin 1
cos
1

sin 21

cos 21

116

The Zoeppritz Equations at


0 degrees
Although the Zoeppritz equations look intimidating, in the case of normal
incidence the equations give us the following simple values for the reflection
and transmission coefficients (see Appendix 1 for the mathematical details):

RS (0o) = RS0 = 0, TS (0o) = TS0 = 0,

2VP2 1VP1
RP(0 ) = RP0 =
,
2VP2 + 1VP1
21VP1
o
TP(0 ) = TP0 =
=1 RP0.
2VP2 + 1VP1
o

These equations tell us that there is no S-wave component at zero angle,


and the reflection and transmission coefficients are related to changes in
the acoustic impedance (P-velocity x density).
1/2/2008

117

The Zero Offset Trace


We can think of the zero angle, or offset,
trace on a P-wave gather as a set of
reflection coefficients corresponding to the
changes in acoustic impedance (density x Pvelocity) at each interface. The equation is
below, and an illustration is on the right.
This is not the complete story, as the
reflectivity is convolved with a wavelet.

RP 0i

Zi
Ri

Zi+1

Z Pi +1 Z Pi
=
,
Z Pi +1 + Z Pi

where :
Z Pi = iVPi = impedance,

= density.
1/2/2008

118

Convolution
Convolution with the seismic wavelet, which can be written mathematically
as S = W*R, is illustrated pictorially below:

=>

W = Wavelet

R = Reflection
Coefficients
1/2/2008

S = Seismic
Trace
119

The Linearized Zoeppritz Equations

Although the Zoeppritz equations give us the exact amplitudes as a function


of angle, the equations themselves do not lend themselves to an intuitive
understanding of the AVO process for angles greater than zero degrees.
For that reason, although modeling should be done with the Zoeppritz
equations, most AVO theory for analyzing real data is based on a linearized
approximation to the Zoeppritz equations initially derived by Bortfeld (1961)
and then refined by Richards and Frasier (1976) and Aki and Richards (1980).
The equations on the next three slides will show various equivalent
formulations of the Aki-Richards equations.
We will then develop an intuitive understanding of these equations using the
rock physics model developed in the first section.

1/2/2008

120

The Aki-Richards Equation


The Aki-Richards equation is a linearized approximation to the Zoeppritz
equations. The initial form of this equation separated the velocity and
density terms:

RP ( ) = a

VP

VS
+b
+c
, where :
VP

VS

1
,
2
2 cos

V 2
2
S
b = 0.5 2 sin ,

VP
a=

VS
c = 4 sin 2 ,
VP

2 + 1
2

, = 2 1 ,

VP 2 + VP1
, VP = VP 2 VP1 ,
2
V + VS 1
, VS = VS 2 VS 1 ,
VS = S 2
2
1 + 2
and =
.
2
VP =

However, in this course we will extensively use the following two


reformulations of the original form given above.
1/2/2008

121

Wiggins Form of the Aki-Richards


Equation
A totally equivalent form of the Aki-Richards equation was derived by
Wiggins et al. (1983). They separated the previous equation into three
reflection terms, each weaker than the previous term:

RP ( ) = A + B sin 2 + C tan 2 sin 2 ,


1 VP
+
where : A =
,

2 V p
2

VS VS
VS
1 VP
=

4
2
,
B

2 Vp
VP VS
VP
1 VP
.
C=
2 Vp
This is often considered to be the AVO equation. A is called the
intercept, B the gradient, and C the curvature. We will therefore refer to
this equation as the ABC equation.
1/2/2008

122

Fattis Formulation of the


Aki-Richards Equation
An alternate form of the Aki-Richards equation was formulated by Fatti et
al. (Geophysics, September, 1994) and is written:

RP ( ) = dRP 0 + eRS 0 + fRD ,


VS2 2
VS2
1
2
where : d = 1 + tan , e = 8 2 sin , f = 2 2 sin tan 2 ,
VP
VP
2
2

RP 0

1 VP
1 VS

, RS 0 =
, and RD =
.
+
=
+

2 VP
2 VS

Notice that the RP0 term given above is identical to the A term in the
previous equation. This equation will be used later in the course as the
basis for pre-stack inversion.
1/2/2008

123

A Summary of the Aki-Richards


Equation
All three forms of the Aki-Richards equation consist of the sum of three
terms, each term consisting of a weight multiplied by an elastic parameter
(i.e. a function of VP , VS or ). Here is a summary:

Equation
Aki-Richards
Wiggins et al.
Fatti et al.

Weights

a , b, c
1, sin 2 , tan 2 sin 2

d , e, f

Elastic Parameters

VP VS
,
,
VP
VS

A, B, C

RP 0 , RS 0 , RD

Note that the weighting terms a, b, c and d, e, f contain the squared VP/VS ratio
as well as trigonometric functions of . However, in the Wiggins et al.
formulation, this term is in the elastic parameter B.
1/2/2008

124

Physical Interpretation
A physical interpretation of the three equations is as follows:
(1) Since the seismic trace consists of changes in impedance rather than
velocity or density independently, the original form of the Aki-Richards
equation is rarely used.
(2) The A, B, C formulation of the Aki-Richards equation is very useful for
extracting empirical information about the AVO effect (i.e. A, which is
called the intercept, B, called the gradient, and C, called the curvature)
which can then be displayed or cross-plotted. As pointed out in the
previous slide, explicit information about the Vp/Vs ratio is not needed
in the weights.
(3) The Fatti et al. formulation gives us a way to extract quantitative
information about the P and S reflectivity which can then be used for
pre-stack inversion. As shown in Appendix 1, the terms RP0 and RS0
are the linearized zero-angle P and S-wave reflection coefficients.
1/2/2008

125

Wet and Gas Models


Let us now see how to get from the geology to the seismic using the
second two forms of the Aki-Richards equation. We will do this by using
the two models shown below. Model A consists of a wet, or brine, sand,
and Model B consists of a gas-saturated sand.

VP1,VS1, 1

VP2,VS2, 2
(a) Wet model
1/2/2008

VP1,VS1, 1

VP2,VS2, 2

(b) Gas model


126

Model Values
In the section on rock physics, we computed values for wet and gas sands
using the Biot-Gassmann equations. Recall that the computed values were:
Wet: VP2 = 2500 m/s, VS2= 1250 m/s, 2 = 2.11 g/cc, 2 = 0.33
Gas: VP2 = 2000 m/s, VS2 = 1310 m/s, 2 = 1.95 g/cc, 2 = 0.12
Values for a typical shale are:
Shale: VP1 = 2250 m/s, VS1 = 1125 m/s, 1 = 2.0 g/cc, 1 = 0.33
This gives us the following values at the top of the sand/shale zone:
2

Wet Sand:

V
VP
VS

= 0.105,
= 0.105,
= 0.054, and S = 0.25
VP
VS

VP
2

VS
VP
VS

Gas Sand:
= 0.118,
= 0.152,
= 0.025, and = 0.328
VP
VS

VP
1/2/2008

127

Parameters for the ABC and Fatti


eqs.
This leads to the following parameters for the ABC and Fatti equations for
the top of the sands (because of symmetry in this example, the base of sand
values are simply these values multiplied by -1):
Gas Sand:

A = 0.071, B = 0.242, C = 0.059


RP 0 = 0.071, RS 0 = 0.063, RD = 0.025

Wet Sand:

A = 0.079, B = 0.079, C = 0.054


RP 0 = 0.079, RS 0 = 0.079, RD = 0.053

Note that A and B have the same polarity for the gas sand and opposite
polarity for the wet sand, whereas RP0 and RS0 have opposite polarity for
the gas sand and the same polarity for the wet sand. The reason for this
will be clear later.
1/2/2008

128

Aki-Richards Values at Various


Angles
Here are computed values for the ABC and Fatti versions of the Aki-Richards
equation at angles of 0, 30 and 60 degrees:
ABC Method
Fatti Method
Angle/
Sand
1st Term 2nd Term 3rd Term 1st Term 2nd Term 3rd Term RP()
0o Gas

-0.071

o
30 Gas

-0.071

-0.060

60o Gas

-0.071

-0.181

0o Wet

0.079

30o Wet

0.079

o
60 Wet

0.079

1/2/2008

-0.071

-0.006

-0.095

-0.042

6x10

-0.133

-0.285

-0.125

0.025

0.079

-0.020

0.005

0.106

-0.040

-0.002

0.064

-0.060

0.119

0.318

-0.119

-0.061

0.138

-0.071
-5

-0.137
-0.385
0.079

129

Summary of the ABC and Fatti


Methods
There was a lot of information in the last slide, but the key points are:
(1) The individual terms in each approach are different, but the sum is
always identical.
(2) For an angle of zero degrees, the second two terms in both methods
are equal to zero, and the scalar on the first term in the Fatti method
is equal to one.
(3) In the ABC method, the first term is always the zero offset reflection
coefficient, but this is true only at zero angle in the Fatti method.
(4) The third term makes less of a contribution to the sum in the Fatti
method than in the ABC method.
The next slides will show the results at all angles.

1/2/2008

130

Zoeppritz vs the ABC Method


Gas Sand
This figure on the right
shows the AVO curves
computed using the
Zoeppritz equations
and the two and three
term ABC equation, for
the gas sand model.
Notice the strong
deviation for the two
term versus three term
sum.

ABC method:
two term

ABC method:
three term
Zoeppritz

Note: On the next four


plots, the curves have
been calculated as a
function of incident
angle and scaled to
average angle.
1/2/2008

131

Zoeppritz vs the ABC Method


Wet Sand
This figure on the
right shows the AVO
curves computed
using the Zoeppritz
equations and the
two and three term
ABC equation, for
the wet sand model.
Again, notice the
strong deviation for
the two term versus
three term sum.

1/2/2008

ABC method:
three term
Zoeppritz

ABC method:
two term

132

Zoeppritz vs the Fatti Method


Gas Sand
This figure on the
right shows the AVO
curves computed
using the Zoeppritz
equations and the
two and three term
Fatti equation, for
the gas sand model.
Notice there is less
deviation between
the two term and
three term sum than
with the ABC
approach.

1/2/2008

Zoeppritz
Fatti method:
two term
Fatti method:
three term

133

Zoeppritz vs the Fatti Method


Wet Sand
This figure on the
right shows the AVO
curves computed
using the Zoeppritz
equations and the two
and three term Fatti
equation, for the wet
sand model.
As in the gas sand
case, there is less
deviation between the
two term and three
term sum than with
the ABC approach.

1/2/2008

Zoeppritz
Fatti method:
two term

Fatti method:
three term

134

Ostranders Paper
Ostrander (1984) was one of the first to write about AVO effects in gas
sands and proposed a simple two-layer model which encased a low
impedance, low Poissons ratio sand, between two higher impedance,
higher Poissons ratio shales.
This model is shown in the next slide.
Ostranders model worked well in the Sacramento valley gas fields.
However, it represents only one type of AVO anomaly (Class 3) and the
others will be discussed in the next section.

1/2/2008

135

Ostranders Model
Notice that the model consists of a low acoustic impedance and Poissons
ratio gas sand encased between two shales, similar to our previous model.

1/2/2008

136

Synthetic from Ostranders Model

(a) Well log responses for the model.


1/2/2008

(b) Synthetic seismic.


137

AVO Curves from Ostranders Model

(a) Response from top of model.


Note that the transmitted P-wave
amplitude is shifted.

(b) Response from base of model.


Note that the transmitted P-wave
amplitude is shifted.

1/2/2008

138

Shueys Equation
Shuey (1985) rewrote the ABC equation using VP, , and . Only the gradient
is different than in the ABC expression:

1 2

+
,
B = A D 2(1 + D )
2

1 (1 )

VP / VP
+ 1
where : D =
, = 2
, and = 2 1.
VP / VP + /
2
The above equation is quite complicated but can be greatly simplified by
assuming that = 1/3 (the same as Vp/Vs=2). This gives:

1 9

B = A D 2(1 + D ) +
= 2.25 A
2
4

This leads to a very intuitive version of the two-term AVO equation:

RP ( ) = A + (2.25 A) sin 2
1/2/2008

139

Shueys Equation
Gas Sand Model
Aki-Richards vs Shuey

Note that the values are


close but, unlike the
previous three forms of the
equation, Shueys version
does not give exactly the
same values.

0.250
0.200
0.150
0.100
Amplitude

This figure shows a


comparison between the
two forms of the
Aki-Richards equation for
the gas sand considered
earlier.

0.050
0.000
-0.050
-0.100
-0.150
-0.200
-0.250
0

10

15

20

25

30

35

40

45

Angle (degrees)

1/2/2008

A-R Top

Shuey Top

A-R Base

Shuey Base
140

Other Single Layer Models


The previous exercise showed us that for a gas sand with a low acoustic
impedance, we can expect absolute amplitude increases with offset at both
the top and bottom of the sand. For the models, we used P and S-wave
velocity.
Another approach is to use the Poissons ratio change as the key
parameter, utilizing the approximate form of Shueys equation to give us an
intuitive understanding of the amplitude change. This equation was:

RP ( ) = A + ( 2.25 A) sin 2 , where A = RP 0


The next figure shows four single-layer boundaries consisting of all
combinations of increasing and decreasing acoustic impedance and
Poissons ratio. Note that the sign of the gradient is generally the same
as the sign of . (This is not true in the case of a Class 4 sand, as we
shall see in a later theory section.)
1/2/2008

141

Four Single Layer Models

1/2/2008

(a) , VP, and all increase.

(b) , VP increase, decreases.

(c) , VP decrease, increases.

(d) , VP, and all decrease.


142

Multi-Layer AVO Modeling

Multi-layer modeling in the AVO program consists first of creating a stack


of N layers, generally using well logs, and defining the thickness, P-wave
velocity, S-wave velocity, and density for each layer.
1/2/2008

143

Multi-Layer AVO Modeling

You must then decide what effects are to be included in the model: primaries
only, converted waves, multiples, or some combination of these.

1/2/2008

144

AVO Modeling Options


There are three main options for the modeling process:
Zoeppritz Primaries only, using the Zoeppritz equations for calculation.
Aki-Richards Primaries only, using the Aki-Richards equations for
calculation.
Elastic Wave Computation of the full elastic wave solution (with optional
anelastic effects), which includes primaries, converted waves, and
multiples.

1/2/2008

145

Elastic Wave Modeling


Elastic Wave Modeling is the exact solution for a plane wave propagating
through a series of layers.
The theory has been available for a number of years and is described in
Kennet(1979, 1980).
The calculation is done in the frequency domain, so the user must specify
a frequency range, which affects the run-time.
Ideally, the modeling should include all effects, such as multiples and
converted waves. In principle, these can be turned off, but that may
produce instability.
Elastic Wave Modeling can model critical and post-critical events.

1/2/2008

146

Zoeppritz Elastic Wave Comparison

Primaries

Zoeppritz

Elastic wave
Modeling

Multiples

Converted Waves

Refractions

Anisotropy (VTI)

Frequency dependent absorption

Higher order move-out

Post-critical events

Fast computation

The following example, taken from a paper by Simmons and Backus (1994),
illustrates the difference between Zoeppritz and Elastic modeling.
1/2/2008

147

The Oil Sand Model

Simmons and Backus used the thin bed oil sand model shown above.
1/2/2008

148

The Possible Modeled Events

Simmons and Backus (1994)


1/2/2008

149

Responses to Various Algorithms


(A) Primaries-only Zoeppritz, (B) + single leg shear, (C) + double-leg shear,
(D) + multiples, (E) Wave equation solution, (F) Linearized approximation.

Primaries only Zoeppritz


+ single leg shear
+ double leg shear
+ multiples
Wave equation
Aki-Richards
Simmons and Backus (1994)
1/2/2008

150

Logs from the Colony Example

1/2/2008

151

Models from the Colony Example

Zoeppritz

1/2/2008

Aki-Richards

Elastic Wave

152

Zoeppritz vs Elastic Wave Summary

EW Modeling models a wider range of effects, including multiples and


converted waves.
Zoeppritz Modeling is much faster than EW Modeling.
For thin layer models, the results may be significantly different.
EW Modeling should always be more accurate than Zoeppritz Modeling, if
the input parameters are correct. However, EW Modeling may be more
sensitive to log editing problems.

1/2/2008

153

Anisotropy and AVO


So far, we have considered only the isotropic case, in which earth
parameters such as velocity do not depend on seismic propagation angle.
In the next few slides, we will discuss anisotropy, in particular the case of
Transverse Isotropy with a vertical symmetry axis, or VTI.
We will then see how anisotropy affects the AVO response.
Finally, we will look at this effect on our original model.

1/2/2008

154

Isotropic vs Anisotropic (VTI)


Velocity
As mentioned, in an isotropic earth P and S-wave velocities are independent
of angle.
VTI velocities depend on angle, as shown below for three different angles:

VP(90o)
VP(45o)
VP(0o)
VTI can be extrinsic, caused by fine layering of the earth, or intrinsic,
caused by particle alignment as in a shale.
1/2/2008

155

Velocities for Weak Anisotropy


Although the equations for full anisotropy are quite complex, Thomsen
(1986) showed that for weakly anisotropic materials the velocities can be
written as follows, where , , and are called Thomsens parameters. Note
that for AVO and converted wave studies, we are only interested in the first
two velocities and constants. Note also that VSV(0o) = VSH(0o):

VP ( ) = VP (0o ) (1 + sin 2 cos 2 + sin 4 )


2
o

V
(0
)
o
2
2
P
VSV ( ) = VSV (0 ) 1 + 2 o ( )sin cos
VSV (0 )

VSH ( ) = VSH (0o ) (1 + sin 2 )


1/2/2008

156

Thomsens Parameters
Thomsens parameters are simply combinations of the differences between
the P and S velocities at 0, 45, and 90 degrees. The following relationships
can be derived quite easily using the velocities in the previous slide:

VP ( 90 o ) VP ( 0 o )
=
VP ( 0 o )

VSH ( 90 o ) VSH ( 0 o )
=
VSH ( 0 o )

VP ( 45 o ) VP ( 0 o )
VP ( 45 o ) VP ( 0 o )
= 4
+ = 4

o
o
V
(
0
)
V
(
0
)
P
P

In the next slide, we will look at VP and VSV as a function of angle for different values of
and . (As mentioned, VSH will not be used in AVO).

1/2/2008

157

Anisotropic P and SV VTI Velocities

(a) VTI medium


with = 0.2
and = 0.2.
1/2/2008

(b) VTI medium


with = 0.1
and = 0.2.

(c) VTI medium


with = 0.2
and = 0.1.
158

Solving for and using the Velocity

VP(90o)= 2600 m/s

600 m/s

VP(45o)= 2225 m/s

225 m/s
VP(0o)= 2000 m/s

VP ( 90 o ) VP ( 0 o )
=
= 0 .3
o
VP ( 0 )

1/2/2008

VP ( 45 o ) VP ( 0 o )
= 4
= 0.45 0.3 = 0.15
o
V
(
0
)
P

159

AVO and Transverse Isotropy


Thomsen (1993) showed that a transversely isotropic term could be added
to the Aki-Richards equation using his weak anisotropic parameters and
, where Ran( ) is the anisotropic AVO response and Ris( ) is the isotropic
AVO response. Ruger (2002) gave the following corrected form of
Thomsens original equation:

2
Ran ( ) = Ris ( ) +
sin +
sin 2 tan 2 ,
2
2
where : = 2 1 and = 2 1

Ran ( ) = A + B +
2

1/2/2008

2
2

+
+
sin
C
sin
tan

160

Typical Values for Delta, Epsilon and Gamma

Typical values for , , and were given by Thomsen (1986). Here are some
representative values from his table:

1/2/2008

Lithology

VP(m/s)

VS(m/s)

rho(g/cc)

epsilon

delta

gamma

sandstone_1

3368

1829

2.50

0.110

-0.035

0.255

sandstone_2

4869

2911

2.50

0.033

0.040

-0.019

calcareous sandstone

5460

3219

2.69

0.000

-0.264

-0.007

immature sandstone

4099

2346

2.45

0.077

0.010

0.066

shale_1

3383

2438

2.35

0.065

0.059

0.071

shale_2

3901

2682

2.64

0.137

-0.012

0.026

mudshale

4529

2703

2.52

0.034

0.211

0.046

clayshale

3794

2074

2.56

0.189

0.204

0.175

silty limestone

4972

2899

2.63

0.056

-0.003

0.067

laminated siltstone

4449

2585

2.57

0.091

0.565

0.046

161

AVO and Transverse Isotropy


Blangy (1997) computed the effect of anisotropy on models of the three
Rutherford-Williams type. Blangys models are shown below, but since he
used Thomsens formulation for the linearized approximation, his figures
have been recomputed in the next slide for the wet and gas cases using
Rugers formulation. The slide after that shows our example.

1/2/2008

162

Transverse Isotropy AVO Effects

= -0.15
= -0.3

Class 1

Class 1
Class 2

Class 2

Class 3
Class 3

(a) Gas sandstone case: Note


that the effect of and is
to increase the AVO effects.
1/2/2008

Isotropic
--- Anisotropic

(b) Wet sandstone case:


Note that the effect of and
is to create apparent AVO
decreases.
163

Anisotropy Applied to Colony


Example
Isotropic vs Anisotropic AVO
Gas Sand Top, = -0.15, = -0.3

Amplitude

0.000
-0.100
-0.200
-0.300
-0.400
0

10

15

20

25

30

35

40

45

Angle (degrees)
R (Isotropic)

1/2/2008

R (Anisotropic)

164

Anisotropic AVO Model Example

In the above display, we have added simple and logs to the sonic
and density logs from the Colony gas sandstone play in Alberta. Notice
that only the gas sand is isotropic.
1/2/2008

165

Anisotropic AVO Synthetics

(a) Isotropic

(b) Anisotropic

(a) (b)

In this display, the synthetic responses for the logs shown in the
previous slide are shown. Note the difference due to anisotropy.
1/2/2008

166

Conclusions
This section introduced the theory of AVO and considered a number of
modeled examples.
Our first modeled example looked at both a wet sand and a gas sand, which
were based on typical values found in a reservoir. As we will see in the
next section, this is the most common response and is called a Class 3
anomaly.
We also found that modeling can be very sensitive to the type of algorithm
used. For thin beds, wave equation modeling is suggested.
Finally, anisotropy should also be considered, since it can have a large
effect on the AVO response.

1/2/2008

167

The Colony Gas Sand


Creating Synthetics

Reading in Seismic Data


Now we will read in the real
seismic data which tie this
well. Click on Data
Manager/Import Data/Open
Seismic/From SEG-Y File:
On the next menu, select the file gathers.sgy as shown below and click on Add >>

Then click
on Next >>.

1/2/2008

169

Select the option to process this


file as a 2D Line, as shown, and
click on Next >>.

On the next page, make sure that you specify that this file does not have X & Y
coordinates numbers in the trace headers:

Click on Next >> on this page and again on the following page to accept the defaults.
You will see a message, warning you that the seismic file needs to be scanned. Click
on Yes to proceed.

1/2/2008

170

When the scanning


completes, the Geometry Grid
page appears:

Click on Ok to accept the


default geometry.

1/2/2008

171

The Well to Seismic Map menu appears:

This menu tells the program how to map the wells in the GEOVIEW database to the
seismic data. Change the menu as shown above, i.e., enter the number 330 for the
CDP. We have now indicated that AVO_WELL is inserted at CDP 330 . Click on Ok on
this menu and the data set will appear plotted in the AVO Analysis window.

1/2/2008

172

Displaying Seismic Data

Here is the resulting


plot of the gathers:

Move the horizontal scroll bar to display the center portion of the line. You will notice
that the final CDP range will be shown at the bottom of the screen and will change as
you move. Move to a center range of approximately CDP 325-335.
1/2/2008

173

Releasing the mouse button releases the bar and will show this display with the
P-wave log inserted at X 330:

You may notice that the


event at a time of 630 ms
is much stronger in the
center of the line than it
was at the left end. In
addition, there is a
noticeable increase in
amplitude in moving from
near offsets to far offsets.
This is the AVO anomaly
which will be investigated
in the following exercises.

1/2/2008

174

Creating a Super Gather


Next, we will create a super gather, or common offset
stack, around the zone of interest. To do this, first
select Process / Super Gather on the vertical menu
bar of the seismic data window, as shown on the right.

This will bring up the Super


Gather menu, shown on the left.
Keep the default parameters and
click on Next >>.

1/2/2008

175

We next get the menu


shown on the right. Default
the Number of offsets to 10,
but change the Size of
Rolling Window to 5 CDPs,
as shown. This will create
super gathers by averaging
over 5 CDP bin locations.
Note also, that we are
creating 10 offsets in each
output bin, spanning the
range from 53 to 647
meters.

1/2/2008

176

Click on Next >> to get the


menu shown on the right,
which defines the output
format. Default this menu
and click on Ok.

1/2/2008

177

The super gather will


now appear in its own
window, as shown on
the right. Note that the
signal-to-noise ratio of
the event at 630 ms
has been improved.

1/2/2008

178

Creating a CDP Stack


We will also create a CDP stack from the
Super Gather. We will use this to extract
a wavelet and correlate the well.
On the window containing the Super
Gather, click on Process / Stack / CDP
Stack:

On the stacking menus, click on


Next and Ok to create the stack.
Note that the anomaly appears at
the well location.

1/2/2008

179

Seismic Displays
We have just created two seismic displays and will be
creating many more. It can be inconvenient to keep too
many windows open, so any of these seismic windows
can be shut down by clicking on File / Exit Window. To
restore any seismic window, simply click on the Seismic /
Seismic Window Data Menu button on either the Seismic
data window or the AVO Modeling Window, as shown on
the right:

The AVO Open Window Data Menu


is shown on the left. To Show,
Close, or Delete any of the seismic
data in the list, simply highlight the
name and click on the appropriate
button.

1/2/2008

180

Seismic data can also be displayed in the AVO Modeling window. This window
now shows a subset of the CDP gathers inserted in the window.

1/2/2008

181

To change the displayed seismic data,


click on the eyeball button:

and select the


Seismic Views tab,
as shown on the
right. Notice that
the displayed
seismic is gathers.

1/2/2008

182

By using a
combination of the
Delete << and Add>>
buttons, replace the
gathers with the
stack, as shown on
the right.
Click on Ok to
change the display.

1/2/2008

183

The new window looks as shown below. Note that the CDP stack is now displayed
instead of the original raw gathers:

1/2/2008

184

Wavelet Extraction
The next step is to correlate the well.
Before doing that, we should extract a wavelet. We will extract a statistical wavelet
from the CDP stack by doing the following steps:

Click on Wavelet / Extract


Wavelet / Statistical:

1/2/2008

185

When the menu appears, confirm that


the Input Volume is cdp_stack and
change the start time to 500 ms, as
shown on the right.

Click Next >> twice, then change the


Wavelet Length to 100 and click on OK.

1/2/2008

186

The extracted wavelet appears in this


window.

Click on the Frequency tab to see the


amplitude and phase spectrum of
this wavelet. Because we have used
the Statistical option, the phase is
automatically zero-phase.
1/2/2008

187

Log Correlation
Now, we will do Log Correlation. This is important
because, even though we have applied a check shot
correction, there may still be some residual error in
the depth-time curve of the well.
To start that, click on Logs / Correlate:

On the menu, confirm that we are


using the CDP Stack to extract the
composite trace. This composite
trace will be compared with the zerooffset synthetic.
Click on Ok.

1/2/2008

188

The Modeling window now


looks like this.
Note that the blue traces
are the zero-offset
synthetic, while the red
traces show the composite
real trace.
In this step, we try to match
these as closely as
possible, using wavelet
extraction and log
stretching.

1/2/2008

189

Wavelet Extraction using the Well


The previous wavelet extraction did not use the well. For this reason, we can expect
the derived wavelet to have a good estimate of the amplitude spectrum, but not of the
phase spectrum.
Now, we will extract a second wavelet, using the well.

Click on Wavelet / Extract


Wavelet / Use Well:

1/2/2008

190

On the menu pages, confirm


that the parameters are as
shown. Click Next to move
through the menu pages.

1/2/2008

191

Set the time extraction window:

Because the analysis


window is so short, set the
wavelet length to 100 ms.
Click Next and Ok to extract
the wavelet.

1/2/2008

192

The extracted wavelet appears in this


window.

Click on the Frequency tab to see the


amplitude and phase spectrum of
this wavelet. Notice that this new
wavelet is not zero phase.

1/2/2008

193

By zooming in the Log Correlation window, we see that we now have a


reasonably good character match, but there is a residual time mis-match.

To investigate that further, click on the Parameters button.


1/2/2008

194

The Cross Correlation window, which appears


now, is used to help confirm that we have a
good tie.

First, ensure that the correlation is being


calculated over the optimum window. Set the
values as shown, click click Apply.

The correlation plot shows a symmetric shape,


indicating that the phase of the wavelet is
correct.
It also suggests a -6 ms time shift should be
applied to the logs.

1/2/2008

195

To apply that shift, click on the Apply


Shift button on the Log Correlation
window.

The window now


shows a very good
correlation. Click on
Ok to accept this
change.

1/2/2008

196

A menu appears, confirming that we will create


a new sonic log, called P-wave_corr, whose
depth-time curve now incorporates the desired
shift.
Click on Ok to accept that.

We are now ready to create AVO synthetics.


Before doing that, replace the CDP stack in the
display by clicking on the eyeball icon, and
making the changes shown below:

1/2/2008

197

Creating a Synthetic Gather


Now we will make an offset synthetic.

Click on Synthetic /
Zoeppritz:

Set the parameters as shown on the


right. Note that the Target zone has
been set from 400 to 800 meters. The
Zoeppritz equations will be applied only
within this zone.
Also, the Offset range is identical to that
of the Super Gather computed earlier.

1/2/2008

198

Click on Next >> and OK to display the synthetic in the AVO Modeling window:

1/2/2008

199

Now make a second synthetic, using


Elastic Wave Modeling.
Click on Synthetic /
Elastic Wave:

Set the parameters as shown on the


right. Note that the only change from
the previous menu is to set a new name.
Click on Next and accept the defaults on
the next menu page.

1/2/2008

200

When the synthetic calculation is finished, it appears in the window:

Notice that, by default, AVO shows only the last synthetic created. We will change the
layout by clicking on the eyeball icon:
1/2/2008

201

In the View Parameter


menu, go to the Seismic
Views tab, as shown on
the right.
Using the Add>> button,
set the list of Seismic
Views as shown.
Then, highlight each of
the synthetic names,
Elastic Wave Model and
Zoeppritz Model in turn
and set its Excursion to
0.7. This will cause the
overall amplitude to
match the seismic better.
Then, click on Ok to apply
the changes.

1/2/2008

202

The new window should look like this:

Note that two synthetics are similar, and both show the AVO behavior,
which can also be observed on the seismic data. The next step in our
analysis will be to analyze the seismic further, through the use of AVO
attributes.
1/2/2008

203

AVO Analysis on Seismic


Data

Introduction

In the last section, we looked at the theory of AVO and used the equations to
perform forward modeling.
We will now use the linearized Aki-Richards equation to extract intercept and
gradient attributes from seismic data.
We will then look at the classification scheme of Rutherford and Williams,
and discuss AVO cross-plotting.
Finally, we will discuss the use of the third term in the Aki-Richards equation
to derive density attributes.

1/2/2008

205

The Two-Term Aki-Richards Equation


Intercept / gradient analysis is done with the two-term Aki-Richards equation.
Recall that:

R( ) = A + B sin 2
where we have dropped the C term and define A and B as:

1 VP
A=
+
,

2 V p

VS VS
VS
1 VP
B=
,
4
2
2 Vp
VP VS
VP

or, using Shueys approximation:

VP / VP
1 2

+
=
B = A D 2( 1 + D )
,
D
.
2

V P / V P + /
1 (1 )

1/2/2008

206

The Two-Term Aki-Richards Equation


It is common practice to use only 2 terms because:
(1) It simplifies the analysis considerably
(2) For angles less than about 40 degrees, the third term is not significant, as
shown previously:

1/2/2008

207

Estimating the Intercept and


Gradient
Our objective is to determine values for A and B by analyzing real seismic
data.
Step 1 involves converting from offset to angle.
Step 2 involves fitting a regression line to the amplitude picks as a function
of the sine of the angle squared.
The next two slides will illustrate this in simple fashion.

1/2/2008

208

Converting from Offset to Angle


450

The offset
domain is the
conventional
CDP stack
with each trace
at a different
offset. The
acquisition
geometry is
shown below.

1/2/2008

Offset (m)

6000

Angle (degrees)

90

The angle domain


represents a
theoretical
acquisition
geometry in which
each trace
corresponds to a
constant incidence
angle.

209

Conversion from offset to angle can be done using one of these options:
(1) Straight ray assumption (constant velocity)
(2) Ray-tracing (variable velocity)
(3) Ray Parameter approximation (variable velocity approximation)

(1 ) Straight Ray :
X
X
tan =
=
,
2 d Vt 0
where X = offset ,
VRMS t 0
,
2
t0 = 2 way time,
d = depth =

( 2) Ray Parameter :
XVINT
sin =
,
2
tVRMS
where VINT = Interval velocity ,
t = total traveltime.

VRMS = RMS velocity .

1/2/2008

210

Ray
Tracing

Ray
Parameter

The Ray Parameter method is


usually used when analyzing
seismic data.
This is because it is much faster
than Ray Tracing.
Because of the approximations, it
begins to degrade slightly at
angles greater than 50 degrees.
The example on the right shows
angle contours calculated using
Ray Tracing and Ray Parameter.

1/2/2008

65

35 43

50

65

35 43

50

211

Real Data Example


Now lets look at some real data, and see if it matches the theory.
The next slide shows the logs in the discovery well.
After that, we will look at the stack, a classic bright spot.
The slide following the bright spot shows a group of 2D gathers over a gas
zone, in WTVA and color amplitude envelope. Notice the increase in
amplitude as a function of offset.
The slide after that shows a common offset stack or super-gather over
the gathers. The amplitudes have been picked and displayed, to quantify the
amplitude increase.
We will finally discuss the A and B attributes and their combinations.

1/2/2008

212

Logs

The well logs are from the gas discovery well are shown above. Notice
that there is a cleaner response on the density log than the sonic log.
1/2/2008

213

Stack

Bright Spot

The CDP stack over the gas well is shown above, with the sonic log from the
gas well spliced in. Note the bright spot, which may or may not indicate
gas (it could be lithology induced).
1/2/2008

214

Seismic Gathers over a Gas Sand


(a) A series
of corrected
CDP gathers
over a gas
zone.

(b) The same


gathers, but
shown with
color
amplitude
envelope.

1/2/2008

215

Common Offset Stack from Gathers


(a) Common offset stack
(b) Picks from
the trough.

(c) Picks from


the peak.

1/2/2008

216

Common Offset Picks as


Function of sin2
The pick amplitudes are extracted at all
times, two of which are shown.

Offset

+A

+B
sin2

Time
The Aki-Richards equation predicts a
linear relationship between these
amplitudes and sin2.

-A

-B

Regression curves are then


calculated, to give A and B values for
each time sample.
1/2/2008

217

Intercept: A
The result of this
calculation is to
produce 2 basic
attribute volumes

Gradient: B

1/2/2008

218

Derived Attributes
The raw A and B attribute volumes are rarely used in that form. Instead,
other AVO attributes are usually calculated from them.

The most popular AVO attributes are:


(1)
(2)
(3)
(4)

1/2/2008

AVO Product : A*B


Scaled Poissons Ratio Change : A+B
Shear Reflectivity : A-B
Fluid factor

219

Derived Attributes AVO Product :


A*B
Many AVO anomalies have the form
shown at the right.
In this case, both the intercept (A) and
the gradient (B) are large numbers or
bright. Also, they have the same
sign.

+A

+B

This is an example of a Class 3


anomaly.
Forming the product of A and B, we
get:

sin2
-A

-B

Top of sand : (-A)*(-B) = +AB


Base of sand : (+A)*(+B) = +AB
This gives a positive bright response
at both top and base.
1/2/2008

220

The AVO product shows a positive response at the top and base of the
reservoir:

Top

Base

1/2/2008

221

Derived Attributes
Scaled Poissons Ratio Change : A+B
The second combination is derived from Shueys equation:
R( ) = A + B sin 2
A=

1 VP
+


2 V p

1 2

B = A D 2( 1 + D )
+
1 ( 1 )2

If we assume the background Poissons Ratio = = 1/3, then:

B = A D 2(1 + D ) +
=
A
2
2
(2
/
3)
4

or:
1/2/2008

A+ B =

The sum A+B is proportional to the


change in Poissons Ratio.
222

Derived Attributes
Scaled Poissons Ratio Change : A+B
The AVO sum (A+B) shows a negative response at the top of the reservoir
(decrease in ) and a positive response at the base (increase in ):

Top

Base

1/2/2008

223

Derived Attributes
Shear Reflectivity : A-B
The third combination is derived from the Aki_Richards equation:
R( ) = A + B sin 2

1 VP
A=
+

2 V p

2

V VS
V
1 VP
4 S
2 S
B=
2 Vp
VP VS
VP

If we assume the background VS /VP = 1/2, then:


2

V VS
V
1 VP
B=
4 S
2 S
2 Vp
VP VS
VP
= RP 2 RS , where : RP = A, and RS =
or:
1/2/2008

RS =

1
(A B)
2

1 VS
+

2 VS

The difference A-B is proportional to


the Shear Reflectivity.
224

A-B Difference Attribute


The AVO difference (A-B) shows an increase in Shear Impedance at the top
of the reservoir. This calculation is usually done with the more accurate
Fatti equation, which we will see next.

Top

Base

1/2/2008

225

RP and RS Attributes
An alternate form of the Aki-Richards equation was formulated by Fatti et al.
(Geophysics, September, 1994) which can be written (for 2 terms) as:

R ( ) = c1RP + c2 RS
VS
where : c1 = 1 + tan , c2 = 8 sin , =
VP
2

1 VP
1 VS
RP =
, RS =
.
+
+

2 VP

2 VS

This allows us to calculate RP and RS volumes from seismic data in exactly
the same way as A and B volumes.

1/2/2008

226

RP and RS Attributes
RP

RS

1/2/2008

227

RP and RS Attributes

The RP/RS attributes are usually transformed into one of 2 new attributes:
(1) Fluid Factor, which will be discussed next.
(2) AVO Inversion or Lambda-Mu-Rho, which will be discussed in a
later section.

1/2/2008

228

Derived Fluid Factor Attribute


The Fluid Factor attribute (Smith and Gidlow, 1987, Fatti et al., 1994) is
based on Castagnas mudrock equation, which is assumed to be true for
non-hydrocarbon filled layers:

VP = 1.16 VS + 1360 m / s
From this equation, we can derive the following equivalent equation:

RP = 1.16

VS
RS
VP

The Fluid Factor is defined to highlight layers where Castagnas equation


is not true, i.e., potential hydrocarbon zones:

VS
F = RP 1.16
RS
VP
Note that the factor 1.16
1/2/2008

VS
is often customized to fit the local data.
VP
229

Derived Fluid Factor Attribute


The AVO Fluid Factor shows strong deviation from Castagnas equation at
both the top and base of the reservoir. The false bright spot from the coal
is still visible however, indicating that we could do better. This is probably
because the scaling in the fluid factor is not quite right due to the proximity
of the two anomalies.

Top

Base

1/2/2008

230

Important Parameter Choices for


Gradient Calculations

Velocities for
angle calculation
Range of
angles

Minimum Correlation
for handling noisy
picks.

Type of analysis

1/2/2008

231

The Colony Gas Sand


Calculating AVO Attributes

Creating Angle Gathers


First, we will calculate angle gathers from a real data set and observe the effect of the
input velocity on the result. The super gathers are displayed in an AVO seismic
window and look like this:

Anomaly

This data set is a set of CDP-gathered data with a large AVO anomaly. In order to do
AVO analysis, we must be sure that there is a good range of incidence angles at the
zone of interest. To see that, we will convert the input data set to the Angle Gather
domain.
1/2/2008

233

Click on Process / Angle Gather:

On the first page of the menu, we


will keep the default parameters.
Click on Next >>:

1/2/2008

234

On this page, we must


specify the velocity
field which the
program will use to
convert from offset to
angle. We will use the
sonic log. Click on the
Open Well Log button:

1/2/2008

235

Select the log to use for the velocities in the angle transform calculation:

Fill in the menu as shown above, and click on Ok.


1/2/2008

236

The previous menu will now look like this:

Click on Next >> and OK to


produce the result.
1/2/2008

237

The range of angles at any time is a function of the velocity field input. This
calculation shows that at the time of interest, 630 ms, we have angles out to about 30
degrees. This should be good enough for the subsequent analysis. We will use this
angle_gather volume later when we do simultaneous inversion. For now, you can
close it by clicking on File/Exit Window on this window.
1/2/2008

238

Now, we will calculate the AVO attributes and observe the effects of changing
parameters on the result. The data for this exercise is the set of super gathers in the
AVO seismic window, which should look like this:

1/2/2008

239

First, we will pick the anomalous event. To do that, click on


Horizon / Pick Horizons:
On the first menu, click on Ok to
accept the default name:

Modify the two parameters


shown below, which will allow
you to pick a trough on all traces
of the line:

1/2/2008

240

After you have modified the menu as shown on the previous slide, click near the
trough at about 630 ms. This will cause the entire event to be picked:

When you are satisfied that the proper event has been picked, click on Ok to accept
the picks.
1/2/2008

241

To display the pick amplitudes, click on AVO Analysis / Pick


Analysis:

On the first page of this


menu, accept the option
to Show Pick Amplitude
Values Only:

On the second page,


ensure that the
correct horizon is
selected:
1/2/2008

242

After clicking OK
on the menu, the
picks appear at
the base of the
display.

To remove the amplitude display, click on AVO Analysis / View Pick Analysis, and
this will toggle the display off.
1/2/2008

243

Now we will calculate AVO Attribute volumes.


To produce the intercept and
gradient volumes for this data set,
choose AVO Attribute / AVO
Attribute Volume on the window
containing the super gathers.

We will keep all the defaults on the


first page, so click on Next >>:

1/2/2008

244

Take a look at all the default menu items.


Use the log for
the velocities.

We will use only


the first two
terms in the AkiRichards
equation.

Click on Next >>


and OK to start
the calculation.

1/2/2008

245

Here is the resulting AVO attribute calculation:

Notice that there is a strong red anomaly in the center of the section. As shown by
the Color Key, red corresponds to a large positive amplitude. As indicated at the top
of the display, the product of intercept and gradient (A*B) is being displayed in color.
1/2/2008

246

Actually, the horizon display is obscuring the top of the gas


sand event. Lets remove that horizon temporarily. First
click the eyeball icon:

On the parameters menu,


select the Horizons page:

Then, select the option


Do Not Show, and click
on Ok:

1/2/2008

247

Now, we can clearly see the red (positive) response at the top and base of the gas sand,
which is expected for a Class 3 AVO anomaly:

1/2/2008

248

Another, very useful AVO attribute is the Scaled Poissons


Ratio, A+B. To see that, click the eyeball icon once again:

On the first, General, page,


change the Color Data Volume
item as shown, and click Ok:

1/2/2008

249

The result will look like this:

Once again, we see the expected response for the gas sand a negative (orange)
change in Poissons Ratio at the top and a positive (yellow) change at the base of
the layer.
1/2/2008

250

Cross-Plotting AVO
Attributes

AVO Cross-Plotting
AVO cross-plotting involves plotting the intercept against the gradient and identifying
anomalies. The theory of cross-plotting was developed by Castagna et al (TLE, 1997,
Geophysics, 1998) and Verm and Hilterman (TLE, 1995) and is based on two ideas:
(1) The Rutherford / Williams Classification Scheme.
(2) The Mudrock Line.

Rutherford/Williams Classification
Rutherford and Williams (1989) derived the following classification scheme for AVO
anomalies, with further modifications by Ross and Kinman (1995) and Castagna (1997):
Class 1:
Class 2:
Class 2p:
Class 3:
Class 4:

1/2/2008

High impedance sand with decreasing AVO


Near-zero impedance contrast
Same as 2, with polarity change
Low impedance sand with increasing AVO
Low impedance sand with decreasing AVO

252

Rutherford/Williams Classification
The classification depends only on the
contrast in Acoustic Impedance between
the target sand and the surrounding
shales:

Acoustic Impedance
= VP

Shale

Sand

Shale
Class 1

1/2/2008

Class 2

Class 3

253

Rutherford/Williams Classification
These are the generic AVO curves at the top of the gas sand:

1/2/2008

254

An Example of a Class 1 Anomaly


(a) Data example.

(b) Model
example.

Rutherford and
Williams (1989)
1/2/2008

255

Angle Stacks over Class 2 & 3 Sands


(a) Class 2 sand.

(b) Class 3 sand.

Rutherford and
Williams (1989)
1/2/2008

256

Class 2p vs Class 2 Sands


For known Class 2 and 2p anomalies, certain processing options may
enhance the interpretation:

Ross and Kinman (1995) suggest creating a near trace range stack (NTS)
and a far trace range stack (FTS).
For Class 2p: Final Stack = FTS - NTS
For Class 2: Final Stack = FTS
1/2/2008

257

Class 2 Sands
(a) Full stack of a class 2
sand.

(b) FTS of a class 2 sand.

Ross and Kinman (1995)


1/2/2008

258

Class 2p Sands
(a) Full stack of a class
2p sand.

(a) FTS - NTS of a class


2p sand.

Ross and Kinman (1995)


1/2/2008

259

Class 4 Anomalies
Castagna (1995) suggested that for a very large value of A, and a small
change in Poissons ratio, we may see a reversal of the standard Class 3
anomaly, as shown below. Castagna termed this a Class 4 anomaly. Here is
a simple example using Shueys approximation:

9
B = A,
4
(1) If = 0.3 and A = 0.1, then B = -0.575 (Class 3)
(2) If = 0.1 and A = 0.3, then B = 0.075 (Class 4)

1/2/2008

260

Here is Figure 7 from Castagna


et al (1998), which illustrates
the concept of the Class 4
anomaly in more detail.

1/2/2008

261

The Mudrock Line


The mudrock line is a linear relationship between VP and VS derived by
Castagna et al (1985). The equation is as follows and the plot from their
original paper is shown below:

VP =1.16 VS+1360 m/sec

1/2/2008

262

Intercept vs Gradient Crossplot


By using the Aki-Richards equation, Gardners equation, and the ARCO
mudrock line, we can derive a relationship between intercept and gradient.
Recall that:

1 VP
A=
+

2 V p

VS
VS VS
1 VP
B=
,
2
4
2 Vp
VP
VP VS

1 VP
=
Gardner :
4 VP
If we assume that VP = cVS, then we can derive the following relationship:

4
B=
5
1/2/2008

A1 2
c
263

Now let us use a few values of c and see how the previous equation
simplifies. If c = 2, the most commonly accepted value, the gradient is the
negative of the intercept (a -45 degree line on a crossplot):

4
B=
5

9
A1 = A
4

If c = 3, the gradient is zero, a horizontal line on the crossplot of intercept


against gradient:

B=

4
5

9
A1 = 0
9

Various values of c produce the straight lines (wet trends) shown on


intercept / gradient crossplots on the next page.
1/2/2008

264

Mudrock lines on a
crossplot for various
VP/VS ratios (Castagna and
Swan, 1998).

1/2/2008

265

By letting c = 2 for the background wet trend, we can now plot the various
anomalous Rutherford / Williams classes (as extended by Ross and Kinman
and Castagna et al)
Note that each of the classes will plot in a different part of the intercept /
gradient crossplot area.
The anomalies form a rough elliptical trend on the outside of the wet trend.
This is shown in the next figure.

1/2/2008

266

Gradient
Base II P

Base II

Base I
Base III

Top IV

Intercept
Base IV
Top III

Crossplot
Showing
Anomalies
1/2/2008

Top I

Wet Trend
Top II

Top II P

Vp

= 2
Vs

267

ARCO Examples of Cross-Plotting

(a) Cross-plot of well log derived A and B.

(b) Cross-plot of seismically derived A and B.

The scatter on the seismically derived data can be attributed to:


(1) VP/VS Ratio variability
(2) Wavelet interference.
1/2/2008

Foster et al (1993)

268

Intercept / Gradient Cross-Plots

(a) Uninterpreted gas zone


(b) Interpreted gas zone
1/2/2008

269

Seismic Display from A/B Cross-Plots

(a) Before interpretation

(b) After interpretation


1/2/2008

270

Polarization and the AVO Hodogram

We have seen that one of the major causes of scatter on seismically


derived cross-plots is wavelet interference.
One approach to reducing wavelet interference is to use the AVO hodogram
(Keho et al: The AVO hodogram: Using polarization to identify anomalies,
TLE, November, 2001 and Mahob and Castagna: AVO hodograms and
polarization attributes, TLE, January, 2002).
A second approach is to perform inversion of the AVO attribute volumes. We
will see this in a later section.

1/2/2008

271

Up to now, we have
calculated cross plots of A
and B, using fairly large
analysis windows.

This gives a broad cluster of


points, containing both the
background trend and the AVO
anomalies:
1/2/2008

272

In Polarization Analysis, we calculate


cross plots over small sliding windows
on a single trace.

1/2/2008

273

For each cross plot, we


can calculate the
polarization vector.
This measures the
direction of the
dominant energy for
this cluster.

The length of the


vector measures the
average energy in the
cluster.

1/2/2008

274

Theoretically, we can expect wet trend


points to fall around the -45o trend,
while class 3 AVO anomalies fall
around +45o.

-45o

+45o

1/2/2008

275

The Hodogram

time
One way to display this result is to plot the calculated polarization vector on
a 3-D display with time as the third axis. This is called a Hodogram.
1/2/2008

276

A more conventional display shows the


calculated polarization angle for a single
trace as a function of time.
Note that this result depends on the size of
the sliding window.
Also note the anomalous positive
Polarization Angle around 630 ms
indicating the Class 3 anomaly.

1/2/2008

277

Colony Gas Sand Polarization


In addition to the Polarization Angle itself, a very useful attribute is the
Polarization Product, which is Polarization Angle multiplied by the length of
the Polarization Vector. This is expected to highlight bright spots which
have high hydrocarbon potential:

1/2/2008

278

Problems in Intercept / Gradient


Analysis
There are a number of problems that can reduce the accuracy of intercept /
gradient analysis and crossplotting:
Noise on the far offsets
This can be reduced by using a robust, or L1 norm, approach to
fitting the intercept/gradient line.
Misalignment of events at far offsets
This can be reduced by applying a fourth order correction, or by
using a trim static (see the Gulf Coast Exercise).
Neglecting the third term in Aki-Richards
This can be improved by estimating the third (C) term.

1/2/2008

279

Noisy Amplitudes
Corr = .88
In addition to fitting the
regression curves for A and B,
it is important to measure how
well the curves fit the data.
The Gradient Analysis display
shows visually the goodnessof-fit, and displays the
Normalized Correlation, which
measures the goodness-of-fit
on a scale of 0 to 1.

Corr = .45

This correlation can be used


later to reject unreliable data.

1/2/2008

280

Misalignment of Events
One of the most serious problems for Gradient analysis is residual NMO:

Bad for AVO

Good for AVO

This is often most apparent on long-offset data, where high order moveout
corrections are required. Other causes are anisotropy and poor processing.
We will use trim statics to correct this in a later exercise.
1/2/2008

281

Three Term AVO


As mentioned previously, one problem in AVO and crossplot analysis is that
only the first two terms of the Aki-Richards equation are usually extracted
from the CDP gathers. Recall that the full Aki-Richards equation, as shown
below, has three terms:

R( ) = A + B sin 2 + C tan 2 sin 2


where:

1 VP
+
A = RP 0 =

2 VP
VS
VP

4
2
B=

VS
2Vp
2

VS
C=
, and = .
2Vp
VP

VP

1/2/2008

282

An alternate form of the Aki-Richards equation was formulated by Fatti et


al. (Geophysics, September, 1994) which can be written:

R ( ) = c1 RP + c2 RS + c3 RD
where : c1 = 1 + tan 2 , c2 = 8 sin 2 ,
2

VS
1
2
2
c3 = tan 2 sin , = ,
2
VP
1 VP
1 VS

RP =
+
+
.
, RD =
, RS =
2 VP
2 VS

Either the A, B, C and the RP, RS, RD terms can be extracted from the
seismic gathers using a least-squares fitting technique with different
weighting coefficients.
1/2/2008

283

Density Term
Both forms of the Aki-Richards equation can allow us to estimate density
variations.
Using the original A,B,C form, we see that:

1 V VP
AC = P +

2 VP
2Vp 2
This means that if we can estimate all three coefficients, we can generate a
density attribute volume.
That can be very valuable since density is a direct measure of hydrocarbon
saturation. This could solve the fizz water problem.
However, the third coefficient can be very noisy since it depends on the far
angle data (>45 degrees), and is very sensitive to noise.

1/2/2008

284

Gulf of Mexico Example

Top
Base

These are angle gathers from the Gulf of Mexico, showing a strong Class II
AVO anomaly. Angles range from 0 to 60 degrees. The target layer is
annotated at right.
1/2/2008

285

3 Term Gradient Analysis


These displays show the results of fitting the Aki-Richards equation, using 2
and 3 terms, to the event highlighted on the previous slide.
Note that the equation for 2 terms begins to deviate from the seismic picks
after about 45 degrees.
2 Term

3 Term

Base

Top

1/2/2008

286

Poissons Ratio Maps: 2 and 3 Terms

The Scaled Poissons


Ratio maps for the two
results are different,
even though each
depends only on A and
B.
Because of the good
seismic data with
angles to 60 degrees,
we expect the 3-term
result to be more
reliable.
2 Term
1/2/2008

3 Term
287

Elastic Contrasts from 3 Term AVO

Delta Rho

Delta VS

Delta VP

The map of the density term highlights the same areas as the delta VS.

VP/VP is perhaps best in this case: the hydrocarbon anomaly is associated


with a strong change in VP.
1/2/2008

288

Conclusions
This section discussed the AVO intercept and gradient method.
First, we looked at the theory behind the intercept and gradient.
We then looked at an example from a shallow gas sand.
We then discussed the AVO cross-plotting technique.
We then discussed and showed examples of the AVO hodogram method.
Finally, we showed how to estimate three terms from the Aki-Richards
equation and showed several examples of this approach.

1/2/2008

289

Appendix:

Hiltermans Approximation
Hilterman re-arranges Shueys equation in a slightly different way:

R( ) = A + (2.25 A)sin 2
= A (1 sin 2 ) + 2.25 sin 2
= A cos 2 + 2.25 sin 2
Notice that this equation is very intuitive, since it shows that, as the
angle increases, so does the dependence on . Keep in mind that this
equation is strictly correct only for = 1/3 and that the C term has been
dropped. Note also that another way of writing this equation is as
follows, which shows the dependence on A and B:

R( ) = A cos 2 + ( A + B ) sin 2
1/2/2008

290

The Colony Gas Sand


Cross Plotting AVO
Attributes

To perform cross-plot analysis, click on the AVO


Analysis / Cross Plot option in the window containing
the AVO attributes:

On the first page of the menu,


set the CDP range from 300 to
360 and plot A versus B:
Click on Next >> to get the
next page.

1/2/2008

292

Well plot the Intercept along the X-axis and Gradient along the Y-axis. Click Next >>
on these 2 pages.

1/2/2008

293

When you reach this page, you are


setting the time window over which
to extract points for the cross plot.
Note that we are using a 100 ms
window centered around the picked
event Horizon 1.

Also, we are plotting only the


peaks and troughs from the
intercept trace.
When you have filled in the menu
as shown, click on Next >> and OK
to get the cross plot.

1/2/2008

294

AVO cross plot of


intercept and
gradient:

1/2/2008

295

Now we wish to identify points on the cross plot and see


where they come from on the original seismic section. To do
this, click on Zones / Add to get the Add Zone menu.

The menu will look like the one to the


left:
Note that, by default, we are adding a
zone called Zone_1, which will be a
gray colored Polygon

1/2/2008

296

To create the zone, use the


mouse to highlight the
region on the cross plot by
clicking the left mouse
button at each of the
corners of the polygon.
When you have done, click
the right mouse button to
complete the polygon.
Define a zone roughly as
shown to the right:

Click on Apply on the Add Zone menu to add this zone.


1/2/2008

297

Now click on Cross-Section to get a new window with the seismic data plotted.
You will also see the selected points highlighted with the same color:

The region that we have highlighted is the wet trend, which should not contain the
AVO anomaly.
1/2/2008

298

To add a second zone, go back to the menu and click Add new zone, as shown here:

The menu now changes to


indicate that a new zone,
Zone_2, will be created in
yellow.

1/2/2008

299

Now, draw the polygon, like the


one shown on the right.
When drawing a polygon, click
the left mouse button at each of
the corners. Then finally, click
the right mouse button, when
you wish to close the polygon.

1/2/2008

300

When you have drawn the


polygon, click on Apply and your
screen should look something
like this.

1/2/2008

301

Finally, add a third zone, as


shown to the right:

When you have added the


third zone, click on Apply and
Ok on the Add Zone Menu.
Your final cross plot should
look something like this:

1/2/2008

302

The highlighted regions


in quadrants 1 and 3
should correspond to
Class 3 AVO anomalies.
The cross section shows
that this corresponds to
the anomaly we have
been studying so far:

The cross section shows only the portion of data which was used to calculate the
cross plot. Now that we have identified 3 zones, we wish to see them displayed back
on the original seismic data.
1/2/2008

303

To do that, go back to the window containing the intercept and


gradient volumes, and click on the eyeball icon. This will cause the
View / Parameters menu to appear.
On this menu, click on the Color Data Volume item to get a
pullDown list of all the options for color display:

Select the Cross Plot item and click


on OK to see the new plot.

1/2/2008

304

The resulting display shows the Cross Plot zones which you have interpreted from the
Cross Plot. This is available over the entire line and not just the analysis region:

1/2/2008

305

We have now completed this part of the


work in the Colony project. We will
need it for a later exercise, so click on
Project / Save on one of the seismic
windows:

Then, close down the AVO program by


clicking on File / Exit Project on one of the
seismic windows:

1/2/2008

306

AVO Case Study


Onshore Texas Example

AVO Case Study, Onshore Texas


Example

This case study comes from a paper by Mark Gregg and Charles Bukowski
(Leading Edge, November, 2000).

This paper shows a very practical example of the application of AVO to a


mature basin.

1/2/2008

308

The exploration objective was


the elastic Oligocene Vicksburg
formation in South Texas.
This has produced more than 3
trillion ft3 of gas since the
1920s, but not much AVO work
has been reported.
The authors believe the lack of
AVO application comes because
the Vicksburg trend is not a
typical amplitude-supported
play.

1/2/2008

309

The motivation for using AVO


came from results like those
shown on the left.
Using the conventional poststack data, it is difficult to
distinguish Gas from Wet sand
before drilling.
Prior to AVO analysis, the
authors had drilled one
commercial gas well, one noncommercial gas well, and three
dry holes.

1/2/2008

310

These curves from the gas


discovery well show both a
Gas and a Wet zone.
The change in acoustic
impedance is small but the
change in Poissons ratio is
large.
This suggests a class 2
AVO anomaly.

1/2/2008

311

Synthetic modeling
confirmed the expected
class 2 response.

1/2/2008

312

The data was


reprocessed to include
nonhyperbolic
moveout. This turned
out to be critical, as the
figure shows.

1/2/2008

313

A very useful indicator is the


Near and Far Angle Stack.
Note that the Gas sand shows
its brightest response on the
Far Angle stack, as expected
for the class 2 behavior.
The authors used the Far
Angle Stack as the main tool
for searching for new
anomalies.

1/2/2008

314

The authors studied the existing wells and came to these conclusions:
(1) There were about 100 gas wells in the area with cumulative production
> 1 billion ft3.
(2) About of these were associated with class 2 AVO anomalies.
(3) About 65% of the ~70 drilled anomalies were commercial gas
accumulations.
(4) Thicker, better-developed reservoirs produced the most distinctive
anomalies.
(5) Threshold gross reservoir thickness required to produce an anomaly
was about 30-60ft.
(6) Most productive anomalies were at depths of 5,000-10,000 ft.

1/2/2008

315

This is the first drilled


anomaly.
100 ft gross interval
with 72 ft of net pay,
producing initially 3
million ft3 of gas per
day.
Since the anomaly is
not visible on the
conventional stack,
this would not have
been drilled without
the AVO analysis.

1/2/2008

316

A second anomaly was


identified by interpreting the
far-angle stack using
Landmarks Earthcube
software.
This had not been identified
before AVO, because of the
poor quality of the
conventional stack. This was
presumed to be because of
the small acoustic impedance
contrast.
Note that there appear to be
multiple anomalies at the
prospective level.
1/2/2008

317

The drilled well


encountered 2 pay
zones.
Upper zone: gross
thickness of 54 ft,
with 28 ft net pay.
Lower zone: gross
thickness of 214 ft
with 69 ft net pay.
Initial production
rate was 5.3 million
ft3 with estimated
ultimate recovery of
14 billion ft3.
1/2/2008

318

Two more
successful wells
are shown here.

1/2/2008

319

This is an unsuccessful result. The drilling encountered 105 ft of clean,


low-gas-saturated sand at the anomaly.

1/2/2008

320

Results:
(1) Six commercial discoveries.
(2) Two dry holes, caused by low gas saturation.
(3) This is a 75% success rate, dramatically improved from the original
20% success rate.

Authors conclusions:
(1) Know your rocks. Do the modeling.
(2) Look beyond conventional seismic techniques, e.g. AVO.
(3) Low gas saturation remains a pitfall of the AVO method.

1/2/2008

321

Appendix 1: The Zoeppritz Equations


Zoeppritz derived the amplitudes of the reflected and transmitted waves
using the conservation of stress and displacement across the layer
boundary, which gives four equations with four unknowns. Inverting the
matrix form of the Zoeppritz equations gives us the exact amplitudes as a
function of angle:

sin 1
RP(1) cos
1
R ( )
S 1 = sin 2
1
TP(1)


TS (1) cos 21

1/2/2008

cos 1

sin 2

cos 2

sin 1
sin 2
cos 2

2
2VS2VP1
2VS2VP1
VP1

cos 21
cos
2

cos
2

2
1

1VS12VP2
1VS12
VS1

2VP2
2VS2
VS1
sin 21
cos 22
sin 22
1VP1
1VP1
VP1

sin 1
cos
1

sin 21

cos
2

322

The Zoeppritz Equations at 0


degrees
Although the Zoeppritz equations look intimidating, in the case of normal
incidence the equations reduce to the following simple form:

RP(0o) RP0
o 1
RS (0 ) = RS0 = 0
TP(0o) TP0
o
TS (0 ) TS0 1

1
0
VP1
VS1

0
1
0

2VP2
0
1VP1

1
0
2VS2VP1

2
1VS1

0
1

By performing the above matrix inversion, we will see some interesting


features about the zero angle case.
1/2/2008

323

The Zoeppritz Equations at 0


degrees
The matrix inversion can be done by hand when there are so many zeros
(but great care must be taken with the signs!), and we get:

2VP2
1VP1

0
0

2VP2 + 1VP1
2VP2 + 1VP1
0
2
RP0 2VS2
1VS1

0
0
1
R

(
)

V
V

V
+
+
0
S
1 S1
P1 2 S2

= 2 S2 1 S1

1VP1
1VP1
TP0
0
0
0

VP2 + 1VP1

VP2 + 1VP1 1
2
2
T

S0
2

1VS1
1VS1
0
0

(
)

V
V

V
+
+
P1 2 S2
1 S1

2 S2 1 S1
The zero angle reflection and transmission coefficients are therefore:

21VP1
2VP2 1VP1
RS0 = TS0 = 0, RP0 =
, TP0 =
=1 RP0
2VP2 + 1VP1
2VP2 + 1VP1
1/2/2008

324

The Linearized Approximation


A useful approximation to RP0 can be derived by noting that:

2VP 2 1VP1 Z P 2 Z P1 Z P
Z P1 + Z P 2
RP 0=
=

, where Z P =
.
2
2VP 2 + 1VP1 Z P 2 + Z P1 2 Z P
From calculus, we know that:

d ln( Z (t ))
1 dZ (t )
dZ (t )
=
d ln( Z (t )) =
dt
Z (t ) dt
Z (t )
Replacing the derivative d with the difference operator gives:

RP 0

ln Z P ln VP + ln 1 VP

=
=
+

2
2
2 VP

Notice that the above equation is the linearized A or RP0 term in the AkiRichards equation and its various reformulations.
1/2/2008

325

Appendix 2: Extracting Attributes


In the course we have often discussed the need to extract attributes from
the pre-stack seismic gathers. To see how this is done, note that all the
linearized equations we have looked at so far can be written as:

RP ( ) = f1 p1 + f 2 p2 + f 3 p3 ,
where f1 , f 2 , and f 3 are functions of and sometimes VS2 / VP2 ,
and p1, p2 , and p3 are functions of VP ,VS , and .
For example, the ABC equation is:

RP ( ) = A + B sin 2 + C tan 2 sin 2 , where : f1 = 1, f 2 = sin 2 , f 3 = tan 2 sin 2 ,


2

p1 = A =

1/2/2008

V VS
V
VP
VP
VP
+
4 S
2 S
, p2 = B =
, p3 = C =
.
2V p 2
2V p
2V p
VP VS
VP

326

Extracting Attributes
For N traces, where we know the angles, we can write:

RP (1 ) = f1 (1 ) p1 + f 2 (1 ) p2 + f 3 (1 ) p3
RP ( 2 ) = f1 ( 2 ) p1 + f 2 ( 2 ) p2 + f 3 ( 2 ) p3
M
M
M
M
RP ( N ) = f1 ( N ) p1 + f 2 ( N ) p2 + f 3 ( N ) p3
This can be written in matrix form as:

RP (1 ) f1 (1 )
R ( ) f ( )
P 2 = 1 2
M M

R
(

)
P N f1 ( N )
1/2/2008

f 2 (1 )
f 2 ( 2 )
M
f 2 ( 2 )

f 3 (1 )
p1

f 3 ( 2 )
p2
M
p2
f 3 ( N )
327

Extracting Attributes
The previous equation can be written more simply as:

R = MP,
where R is a known vector of N picked reflection coefficients at a constant
time, M is an N x 3 vector of computed values, and P is the unknown vector
containing the parameters to be estimated.
This is an over-constrained problem which has the following solution:

p1
1 0 0
P = p2 = ( M T M + I ) 1 M T R, where I = 0 1 0,

p3
0 0 1
and is a pre - whitening factor.
1/2/2008

328

Extracting ABC Attributes


Let us take the specific case of extracting ABC attributes, for which the
forward problem is:

RP (1 ) 1 sin 2 1
R ( ) 1 sin 2
2
P 2 =
M M
M

2
(

)
R
1
sin
N
P
N

tan 2 1 sin 2 1
A
2
2
tan 2 sin 2
B


M
C
2
2
tan N sin N

2
Since we know that sin i = X iVINT / tVRMS
, we can write :

RP (1 ) 1
R ( )
P 2 = 1
M M


RP ( N ) 1
1/2/2008

(X V
(X V

1 INT

2
)
/ tVRMS

(X V
) (X V

2 INT

2
/ tVRMS

M
2
2
(X NVINT / tVRMS )

(
(
(

)
)
)

2
2
2
2
)
(
)
/
/
/
tV
X
V
tV
1
1 INT
1 INT
RMS
RMS
A
2
2
2
2
2 INT / tVRMS ) / (X 2VINT / tVRMS ) 1
B

M
C
2
2
2
2
(X NVINT / tVRMS ) / (X NVINT / tVRMS ) 1
329

Extracting ABC Attributes


Simplifying the notation we get the following solution:

1 b1
A 1 1 L 1
B = b b L b 1 b2
N
M M
1 2

c c L cN
C
1 2
1 bN

N
bi
=
i =1
N

ci
i =1

ci

i =1
N

bi ci

i =1
N

2
ci

i =1
N

b
i

i =1
N

bi2

i =1
N

b c

1/2/2008

i i

i =1

c1

c2

cN
1

1 1
b b
1 2
c1 c2

RP1
L 1

R
2
P

L bN
M
L cN

R
PN

RPi

i =1
N

2
bi
2
bi RPi , bi = (X iVINT / tVRMS ) , ci = 1
bi 1

i =1
N

ci RPi

i =1
N

330

Extracting RP0, RS0 and RD Attributes


Next, let us take the case of extracting RP0, RS0 and RD attributes, for which
the forward problem is:

RP (1 ) d (1 ) e(1 )
R ( ) d ( ) e( )
2
2
P 2 =
M
M M

R
(
)
P N d ( N ) e( N )

f (1 )
RP 0

f ( 2 )
RS 0 , where :
M
RD
f ( N )

2
2
1
V
V
d (i ) = 1 + tan 2 i , e(i ) = 8 S2 sin 2 i , f (i ) = 2 S2 sin 2 i tan 2 i ,
2
VP
VP

RP 0

1/2/2008

X iVINT

1 VS
1 VP
,
,
and
sin

.
R
=
=
=
+
,
R
=
+
S0
D
i

2
tVRMS

2 VS

2 VP

331

Extracting RP0, RS0 and RD Attributes


Simplifying the notation we get the following solution:

d1
RP 0 d1 d 2 L d N
R = e e L e d 2
2
N
M
S 0 1

f 2 L cN
f
RD
1
d N

1/2/2008

d d e
2
i

i =1
N

i i

i =1
N

d i ei

i =1
N

ei2

i =1
N

d f e f
i i

i =1

i i

i =1

eN

di fi

i =1
N

ei f i

i =1
N

2
fi

i =1
N

f1

f2

f N

e1
e2
M
1

i =1
N

i =1
N

i =1

d1 d 2
e e
2
1
f1 f 2

RP1
L d N

R
2
P

L eN
M
L f N

R
PN

d i RPi

ei RPi

f i RPi

332

You might also like