Está en la página 1de 345

AVO Workshop Part 2

Bali, Indonesia
January 24, 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

Jan 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

Jan 2008

AVO Inversion

Introduction
In this section, we will review the principles of AVO inversion.
First, we will discuss the general theory.
Then, we will discuss four approaches to this problem:
Elastic Impedance
Independent Inversion
LambdaMuRho Analysis
Simultaneous Inversion
Finally, we will perform an inversion exercise using the Colony data.

Jan 2008

Seismic Lithology Estimation

Gathers

Stack
Inversion
Estimate
Z= VP

Traditional methods of seismic lithology estimation involve stack followed


by inversion. This allows for only the estimation of acoustic impedance,
which is not sufficient for inferring fluid content.
Jan 2008

Seismic Lithology Estimation

Gathers

Stack

AVO Analysis

Inversion

AVO 1 AVO 2

...

AVO N

Estimate VP, VS, and

Estimate
Z= VP

The AVO method allows us to use multiple AVO attributes (shown above
as AVO 1 through AVO N) to simultaneously estimate VP, VS, and , thus
inferring fluid and/or lithology.
Jan 2008

Possible Attributes
But which AVO attributes will give us the best estimate of these parameters?
A number of different possibilities have been proposed:

Offset or angle-limited stacks


Elastic Impedance inversion
Intercept and Gradient
Independent extraction of RP and RS reflectivity
Inversion of RP and RS to give ZP and ZS
Lambda-Mu-Rho analysis of ZP and ZS
Simultaneous Inversion to give ZP, ZS, and Density

We have already discussed the Intercept and Gradient method at length.


In this section, we will discuss some of the other methods.

Jan 2008

Range Limited Stacking

Gathers
AVO Analysis
Near Stack

Far Stack

Above, we see a simple flowchart for range-limited stacking. Range-limited


stacking, using constant offsets or constant angles, is very robust. But
how do we interpret the results?

Jan 2008

Range Limited Stacking

(a)

Here are the (a) near angle


(0o-15o) and (b) far angle
(15o-30o) stacks from the
Colony seismic dataset.
Notice that the amplitude of
the bright-spot event at
about 630 ms is stronger on
the far-angle stack than it is
on the near-angle stack. As
we saw earlier, this is a gassand induced bright-spot.

(b)

Jan 2008

10

Cross-Plotting Angle Range Stacks

Here is a crossplot of
the near and far
offset, with several
high amplitude zones
highlighted.

Jan 2008

11

Cross-Plotting Angle Range Stacks


Here are the highlighted zones from the crossplot shown back on the
seismic section. Note that the gas sand zone has been well delineated.

Top GAS
Base GAS
Coal

Jan 2008

12

From Range Limited Stacking to


Elastic Impedance

Range-limited stacking, using constant offsets or constant angles, is very


robust, and avoids misaligned event problems. But what does it mean?

Patrick Connolly, from BP, came up with a novel approach to the


interpretation of range limited stacks, called Elastic Impedance.

Elastic Impedance is based on the Aki-Richards equation, and the next


few slides will develop the concept.

Jan 2008

13

Elastic Impedance Theory


Recall that the Aki-Richards Equation can be written:

R( ) = A + B sin2 + C sin2 tan2


1 VP 1 AI
+
where A =
=
, AI = VP = acoustic impedance,

2 VP
2 AI
2

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

1 VP
and C =
.
2 VP

Connolly (1999) proposed that, analogously to acoustic impedance, we


could define elastic impedance (EI) as:

1 EI 1
R( )
ln (EI )
2 EI
2
Jan 2008

14

Elastic Impedance Theory


If we let

VS
K =
VP

and note that

sin 2 tan 2 = tan 2 sin 2 ,

we can re-arrange the Aki-Richards equation to get:

VS

1 VP
2
2
2
ln (EI ) =
1 + tan
8 K sin +
1 4 K sin
2 VP
VS

If we let K be a constant, we can write:

(1 4 K sin )
1+ tan )
8 K sin

(
ln ( EI ) = ln VP
ln VS
+ ln

2
(1+ tan 2 )
8 K sin 2 (1 4 K sin )

= ln VP
VS

Jan 2008

15

Elastic Impedance Theory


If we then integrate and exponentiate, we get the following form for EI:

EI ( ) = V

(1+ tan 2 ) ( 8 K sin 2 )


P
S

where

(1-4 K sin 2 )

VS2
K= 2
VP

Note that if = 0o, EI reduces to Acoustic Impedance (AI), where:

( )

EI 0o = AI = VP
Jan 2008

16

Elastic Impedance Theory


The preceding equation used all three terms in the Aki-Richards equation.
For angles greater than 300, this equation does not give a straight line fit.
For a higher angle (larger offsets), we use only the first two terms, which
leads to:

EI ( ) = V

(1+ sin 2 ) ( 8 K sin 2 )


P
S

VS
where K =
VP

(1 4 K sin 2 )

Again note that where = 0o, we get:

EI (0 ) = AI = V
o

Jan 2008

P
17

The acoustic impedance model


The model that forms the basis for acoustic impedance inversion is:

Wavelet

Jan 2008

Acoustic
Impedance (AI)
= VP

Zero offset Preflectivity


RP0

Zero offset
seismic trace
S(0o)

18

Acoustic impedance inversion


Acoustic impedance inversion tries to reverse the forward AI model:

Inverse
Wavelet

Zero offset
seismic trace
S(0o)
Jan 2008

Zero offset Preflectivity


RP0

Acoustic
Impedance (AI)
= VP
19

The elastic impedance model


The model that forms the basis for elastic impedance inversion is:

Wavelet

Jan 2008

Elastic
Impedance (EI)
= VPaVSbc

Aki-Richards
reflectivity at
RP()

Seismic trace
at angle
S()

20

Elastic impedance inversion


Elastic impedance inversion tries to reverse the forward EI model:

Inverse
Wavelet

Seismic trace
at angle
S()
Jan 2008

Aki-Richards
reflectivity at
RP()

Elastic
Impedance (EI)
= VPaVSbc
21

Elastic Impedance Effect of Oil


Saturation
The transformation of an AI log from 0 to 30 results in a generally similar
log but with lower absolute values. The apparent acoustic impedance
decreases with an increase in angle. The percentage decrease is greater for
an oil sand than for shale.

Connolly 1999
Jan 2008

22

Elastic Impedance Data Example


The following figure, from Connolly (1999) shows the computed curves for AI
and EI at 30 degrees:

Jan 2008

23

Elastic Impedance Data Example


The following figure, also from Connolly (1999) shows that when we scale the
curves shown on the previous slide, we get a better separation for the oil
sands using EI over AI:

Jan 2008

24

Colony sand case study

In the following set of slides, we will consider a


case study from the Colony sand in Alberta.
As we have seen, this is a 2D example which lends
itself well to AVO analysis.
The analysis was done using the Hampson-Russell
AVO program.
Note the dramatic change in the elastic impedance
response when we invert for Elastic Impedance at
two different angles.

Jan 2008

25

Gas sand case study

The figure above shows the logs after fluid substitution in the gas zone. The EI_Near
log on in blue was created at 7.5o and the EI_Far log in red was created at 22.5o. Note
that the Near < Far outside the gas sand but Far > Near inside the sand.
Jan 2008

26

Gas sand case study


EI_Near

Jan 2008

EI_Far

(a)
(b)
The figure above shows (a) the interpreted crossplot between the near and
far EI logs, and (b) the zones marked on the logs themselves. Notice the
clear indication of the gas sand zone.

27

Gas sand case study

(a)

(b)

Jan 2008

Here are the (a) near


and (b) far angle stacks
from the seismic dataset.
Notice that the amplitude
of the bright-spot event
at about 630 ms is
stronger on the far-angle
stack than it is on the
near-angle stack. As we
saw earlier, this is a gassand induced brightspot.

28

Gas sand case study


Here is the
comparison between
the inversions of the
(a) near-angle stack
and (b) far-angle
stack, using the
elastic impedance
concept. Notice the
decrease in the
elastic impedance
value on the farangle stack.

Jan 2008

(a)

(b)

29

Gas sand case study


The figure on the left shows a crossplot between
the EI at 7.5o, on the horizontal axis, and the EI
at 22.5o, on the vertical axis. The background
trend is the grey ellipse, and the anomaly is the
yellow ellipse. As shown below, the yellow zone
corresponds to the known gas sand.

Jan 2008

30

EI Inversion Steps
Gathers
AVO Analysis
AVO
Program

STRATA
Program

Near angle
stack at 1

Far angle
stack at 2

Invert to EI(1)

Invert to EI(2)

Cross Plot
Here is a flowchart which summarizes how the EI approach can be
implemented using the Hampson-Russell suite of programs.
Jan 2008

31

Extended elastic impedance (EEI)

A more recent implementation of the elastic impedance


process is called extended elastic impedance or EEI.
This method is an improvement of EI for two reasons:
It gets around the scaling issues associated with EI
(i.e. the values get smaller as a function of angle).
It introduces a new parameter, the angle , which can
be correlated with various other physical parameters
such as , , Vp/Vs, etc., depending on its value.
For those interested in this method, we have added an
appendix (Appendix 1) at the end of this section.

Jan 2008

32

Independent AVO Inversion


and
Lambda-Mu-Rho

Independent inversion for P and


S-Impedance
We now turn from elastic impedance inversion to independent P and
S-impedance inversion. Later, we will discuss simultaneous
inversion for P and S-impedance (and optional density) and show
how this approach differs from the independent approach.
Both independent and simultaneous inversion for P and Simpedance will lead us to the lambda-mu-rho approach, but we will
talk about this method before introducing the simultaneous
approach.
As discussed earlier, utilizing the ARCO mudrock line, was given by
Fatti et al (Geophysics, Sept. 1994) and is used in our AVO program.

Jan 2008

34

Extracting RP and RS
We call this method independent inversion, because the first step is to
extract independent estimates of the zero-offset P and S reflectivities, RP0
and RS0 from the seismic gathers.
This is done using the Fatti equation, which was first introduced in the
modeling part of this course:

RPP ( ) = c1RP 0 + c2 RS 0 + c3 RD
where c1 = 1 + tan , c2 =
2

RS 0 =

Jan 2008

8 sin 2
2
sat

1
2 sin 2
2
, c3 = tan
,
2
2
sat

1 VS
=
,
and
R
.
+
D

2 VS

35

Extracting RP and RS
Let us drop the third term from the previous equation and
consider extracting RP and RS from an N-trace gather:
Angle
1
600

t
Time
(ms)

650

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

2
R
(

)
1
tan
N
+
P N
Jan 2008

Each event in the N-trace


angle gather is picked as
shown on the left. We can
then set up the forward
problem as:

8(VS2 / VP2 )sin 2 1

8(VS2 / VP2 )sin 2 2 RP 0


R = MP

RS 0
M

2
2
2
8(VS / VP )sin 3
36

Extracting RP and RS
The previous equation can be solved using a least-squares
approach:

1 0
P = ( M M + I ) M R, where I =
.

0 1
T

The complete solution with = 0 can be written:


N

2
2
(1 + tan i )

R
P0
i =1
=
N
R
S 0 r2
(1 + tan 2 i )sin2 i

i =1
where r 2 = VS2 / VP2 .

Jan 2008

2
2
2
(1 + tan i )sin i
r

i =1
N

2
2
2
(r sin i )
64

i =1
N

2
RPi (1 + tan i )

i =1
,
N

2
2
RPi sin i
8r

i =1

37

Interpreting the mathematics


The previous two slides were quite mathematical, so let us
review the basic concepts descriptively:
The parameters in the P vector, RP0 and RS0, are evaluated
individually for every sample.
If we use the full three term Fatti equation, we can also
estimate the RD term for density. As shown later, this term is
usually less reliable than the first two terms.
The terms in the M matrix involve angle and VS/VP ratio,
which can be estimated using smoothed well log velocities
and Castagnas relationship to get shear-wave velocity.
The terms in the R vector are the picked seismic amplitudes
along a constant time on each gather.
We next perform independent inversion of the two terms.

Jan 2008

38

Inverting RP and RS
Once we have estimates of RP and RS from the AVO
program, we can then proceed to STRATA to invert both
attributes.
Inverting RP will give acoustic impedance ZP = VP, and
inverting RS will give S-wave impedance ZS = VS. This is
shown in the next slide.
These inverted sections can be displayed or cross-plotted.
Note that in our software, these steps can be done while
staying in the same project, meaning that the resulting
volumes are all accessible from the same database.

Jan 2008

39

RP and RS Inversion

Gathers
AVO
Program

AVO Analysis
RP Estimate RS Estimate

STRATA
Program

Invert to ZP

Invert to ZS

Above is shown a flow chart for the inversion procedure.


Jan 2008

40

Independent Inversion Example

Let us next look at an example of independent inversion as


applied to the Colony sand example.
The inversion procedure used in STRATA involves the
following steps:
Insert the appropriate logs at the correct locations.
Correlate the logs.
Pick the major seismic horizons.
Find an optimum wavelet.
Build the starting model for inversion.
Invert the data.
We will now apply this procedure to the RP and RS sections.

Jan 2008

41

RP Section

Here is the RP section, with the correlated P-wave sonic inserted at


the proper location, and three picked horizons. Horizon 2 is picked
on the gas sand trough.
Jan 2008

42

Wavelet
Next, we will extract
a seismic wavelet
using the statistical
option. Our seismic
data is close to
zero phase, so the
extracted zerophase wavelet is
shown on the left.
We may want to
slightly adjust the
wavelet phase.

Jan 2008

43

P-wave Model

Here is the model result, using a single well and the picked
horizons. The model is scaled to P-Impedance.
Jan 2008

44

P-wave Inversion

Here is the final P-wave inversion result. The low impedance just below
Horizon 2 represents the gas sand.
Jan 2008

45

RS Section

Above is displayed the RS section from the AVO analysis. Notice that
the picked horizons from the RP section are still present.

Jan 2008

46

S-Impedance Model

We now have the created S-Impedance model, as shown above. Note


that the new colour key represents S-wave impedance values.
Jan 2008

47

S-wave Inversion

The result of the S-wave inversion is shown above. Notice that the
gas sand below Horizon 2 is now associated with an increase in
impedance.
Jan 2008

48

The LMR Approach


Goodway et al. (1997) proposed a new approach to AVO inversion based on
the Lam parameters and , and density , or Lambda-Mu-Rho (LMR). The
theory is as follows:

+ 2

VP =
and VS =

therefore : Z S2 = ( VS ) 2 =
and : Z P2 = ( VP ) 2 = ( + 2 )
so : = Z P2 2 Z S2

Jan 2008

49

Interpreting Lambda-Rho & Mu-Rho

The original paper by Goodway et al, gives the following physical


interpretation of the lambda () and mu () attributes: The term, or
incompressibility, is sensitive to pore fluid, whereas the term, or
rigidity, is sensitive to the rock matrix.
As we saw in the theory, it is impossible to de-couple the effects of
density from and when extracting this information from seismic
data.
It is therefore most beneficial to cross-plot vs to minimize the
effects of density.

Jan 2008

50

Extracting Lambda-Rho & Mu-Rho

Once we have estimates of ZP and ZS from the AVO and STRATA


programs, we can then use the Trace Maths option (in either AVO
or STRATA) to produce our lambda-rho and mu-rho volumes.
These volumes can be displayed and cross-plotted.
The flowchart for this is shown in the next slide.
The two slides following the flowchart show vs for a BiotGassmann analysis of a gas sand. Note the vertical separation
for the LMR approach.

Jan 2008

51

LMR Analysis

Gathers
AVO Analysis
RP Estimate RS Estimate
Invert to ZP

Invert to ZS

Transform to and
Cross-plot
Jan 2008

52

The Power of the LMR Method

VP/VS

(VP/VS)2

+2

Shale

2.25

5.1

0.38

20.37

4.035

12.3

3.1

Gas Sand

1.71

2.9

0.24

18.53

6.314

5.9

0.9

Change /
Average (%)

-27

-55

-45

-9.5

44

-70

-110

Goodway et al. (1997)


Petrophysical analysis indicating that / is the most sensitive to variations
in rock properties going from shale into gas sand.

Jan 2008

53

LMR vs Zp/Zs on Well Logs

Goodway et al. (1997)


Jan 2008

54

Lambda-Mu-Rho Cross-Plot

The interpretation of P-impedance vs S-impedance and


lambda-mu-rho crossplot from logs. Note the improvement
for lambda-mu-rho.
Jan 2008

Goodway et al. (1997)

55

LambdaMuRho Observations

The interpretation of P-impedance vs S-impedance and


lambda-mu-rho crossplot from seismic data. Again, note the
improvement for lambda-mu-rho.
From Goodway et al., 1999
Jan 2008

56

Colony sand example

Next, we will apply the LMR method to the Colony seismic


example that we were evaluating earlier.
As input to the process, we will use the independently
inverted P and S-impedances sections shown in the
previous section.

Jan 2008

57

Colony Sand - lambda-rho

The extraction of the lambda-rho section using the ZP and ZS


inverted sections shown earlier.
Jan 2008

58

Colony Sand - mu-rho

The extraction of the mu-rho section using the ZS inverted section


shown earlier.
Jan 2008

59

mu-rho

Colony Sand cross-plot

lambda-rho
A cross plot of the lambda-rho and mu-rho sections. Two zones are
shown, where red=gas and blue=non-gas.
Jan 2008

60

Colony Sand gas sand zone

The interpreted zones from the previous cross-plot, shown now on


the seismic section. Note the continuity of the gas sand in red.
Jan 2008

61

The generalized fluid approach


In Appendix 2, we show that lambda-mu-rho is a particular case of a
more general approach that is rooted in Biot-Gassmann theory.
The key equation is the generalized method is:
2

VP
f = Z cZ = Z Z S2
VP dry
2
P

2
S

2
P

When the dry rock velocity ratio squared is equal to 2, the method
reduces to lambda-mu-rho, but also implies a dry rock Poissons
ratio of 0.
A more realistic value of c is 2.333, which implies a clean, porous
sandstone.
Even better, a value for c should be estimated from the local
geology. Dillon et al. (TLE, October, 2003) found values as high as
2.8 for sediments in offshore Brazil.

Jan 2008

62

Appendix 1: Extended Elastic


Impedance (EEI)
One of the problems with EI is the fact that the values do
not scale correctly for different angles.
This is due to the variable dimensionality found by raising
the velocity and density terms to increasing powers.
Whitcombe (Elastic impedance normalization,
Geophysics, January-February, 2002), proposed solving
this by scaling the EI equation as follows:

V 1+sin
EI ( ) = VP 0 0 P
VP 0

VS

VS 0

8 K sin 2

1+ 4 K sin 2

where VP 0 , VS 0 , and 0 are reference constants.


Jan 2008

63

Extended Elastic Impedance


Whitcombe et al. (Extended elastic impedance,
Geophysics, January-February, 2002), introduced
Extended Elastic Impedance, or EEI, which extended the
normalization approach.
First, they replaced the sin2 term in the standard twoterm Aki-Richards equation with tan, to give the
following expression between :

A cos + B sin
R = A + B tan R =
cos
The authors then introduced scaled reflectivity, Rsc:

Rsc = R cos Rsc = A cos + B sin


Jan 2008

64

Extended Elastic Impedance


This leads to the extended elastic impedance expression,
which involves substituting the scaled reflectivity term into
the scaled elastic impedance expression, to give:

V p V q r
EEI ( ) = VP 0 0 P S ,
VP 0 VS 0 0
where p = cos + sin ,
q = 8K sin ,
r = cos 4 K sin .
Note that this gives the same values as normalized elastic
impedance, but has an alternate interpretation as given on
the next slide.
Jan 2008

65

Extended Elastic Impedance


Since scaled reflectivity goes from A at = 0o to at = 90o,
will go from acoustic impedance at = 0o to what can be
called gradient impedance (GI) at = 90o.
Thus, another form of EEI can be written:

AI cos GI sin
,

EEI ( ) = AI 0
AI 0 AI 0
where AI = VP , AI 0 = 0VP 0 ,
and

Jan 2008

V V 8 K 4 K
GI = AI 0 P S .
VP 0 VS 0 0
66

Extended Elastic Impedance

(a)

(b)

Figure (a) shows EEI functions for various well logs at different angles and
figure (b) compares elastic parameters and their equivalent EEI curves.
Jan 2008

Whitcombe et al. (2002)

67

Extended Elastic Impedance

(a)

(b)

The authors also define lithological impedance, which lies at a value of


o
o
= -51.3 and fluid impedance, which lies at a value of = 12.4 , as
shown in the example above.
Jan 2008

Whitcombe et al. (2002)

68

Appendix 2 Extended lambda-mu-rho

The velocity equations were derived for solid, isotropic


rocks, in which case the two formulations ( and
) are identical.
A key question is: when we turn our attention to the
porous reservoir rock, which term is more applicable,
or K?
As we will shortly see, it doesnt matter, but each term
needs to be expanded for porous media.
The theory for this was developed independently by Biot
(1941), who used and , and Gassmann (1951), who
used K and . It is best described by Krief et al (1990).

Jan 2008

69

Biot theory for porous rocks


Biot (1941) linked the saturated and dry frame to the Lame
coefficients in the following way:

sat = dry + M
2

sat = the Lam coefficient for the saturated rock,


dry = the Lam coefficient for the dry frame,
= the Biot coefficient, or the ratio of the volume
change in the fluid to the volume change in the
formation when hydraulic pressure is constant,
= the modulus, or the pressure needed to force
water into the formation without changing the
volume.

Jan 2008

70

Gassmann theory for porous rocks


Gassmann (1951) linked the saturated and dry frame to the Bulk
modulus in the following way:

K sat = K dry + 2 M
K sat = the bulk modulus of the saturated rock,
Kdry = the bulk modulus of the dry frame,

= the Biot coefficient,


M = the modulus.

Jan 2008

71

Equating Biot and Gassmann


If we equate the two equations proposed by Biot and Gassmann,
we get:

K sat sat = K dry dry


or : sat = dry

Jan 2008

That is, the shear modulus is independent of the fluid content of the
saturated rock.
Also, the second term, 2M, is independent of K and .

72

Biot-Gassmann summary
In summary, we can rewrite the velocity equations in the following way
using the Biot-Gassmann equations:

VP _ sat

K sat + 4 3
sat + 2
=
=
sat
sat

VS =
sat

where :

= dry = sat ,
K sat = K dry + M ,
2

sat = dry + 2 M .
Jan 2008

73

New equation for P-wave velocity

Thus, we can rewrite the equation for P-wave velocity as:

VP =

f +s

f = fluid/porosity term = 2 M
s = dry skeleton term
4
= K dry + = dry + 2
3
Jan 2008

74

Extracting the fluid term


Using the impedances from AVO and inversion, we can thus extract
the product of density times the fluid term, as shown below:

f = ( VP ) 2 c( VS ) 2 = ( f + s c )
That is, the constant c must be chosen so that the term s c is equal
to zero. Using the two approaches:

dry
(1) s = dry + 2 c =
+2

( 2) s = K dry

Jan 2008

K dry 4
4
+ c=
+
3

3
75

The constant term c and s


Note that the constant term c is simply the square of the ratio between
the dry rock P-wave velocity and the dry rock S-wave velocity:

K dry 4
dry
VP
+2=
+
c= =

3
VS dry
2

Jan 2008

Note that the term s is simply given by c(VS)2.

The key question is: how do we find the value of c?

76

(A) From the dry rock Poissons ratio

Note that the dry rock Poissons ratio can be given in terms of c as
shown below:
2

dry

Jan 2008

VP
c2
=
, where c =
2c 2
VS dry

Thus, by estimating or measuring the dry rock Poissons ratio, we


can get an estimate of c.
It is generally accepted that the dry rock Poissons ratio falls
between the values of 0.09 and 0.2.

77

(B) From lab measurements


Murphy et al (1993) measured values of Kdry and for clean quartz
sandstones, and found a value of 0.9 for their ratio, as shown below:

Jan 2008

78

Table of values
Here is a table of values for the various ratios:

c=

2
(Vp/Vs)^2
dry
4.000
3.333
3.000
2.500
(3) 2.333
2.250
2.233
(2) 2.000
(1) 1.333

Vp/Vs dry
2.000
1.826
1.732
1.581
1.528
1.500
1.494
1.414
1.155

dry
0.333
0.286
0.250
0.167
0.125
0.100
0.095
0.000
-1.000

Kdry/
2.667
2.000
1.667
1.167
1.000
0.917
0.900
0.667
0.000

dry/
2.000
1.333
1.000
0.500
0.333
0.250
0.233
0.000
-0.667

Note in the above table that (1) corresponds to K-, (2) to and (3)
to a poroelastic clean sand.
Jan 2008

79

Summary
This appendix has shown that the lambda-mu-rho approach is a
particular case of a more general approach that is rooted in BiotGassmann theory.
The key equation is:
2

VP
f = Z cZ = Z Z S2
VP dry
2
P

2
S

2
P

When the dry rock velocity ratio squared is equal to 2, the method
reduces to lambda-mu-rho, but also implies a dry rock Poissons
ratio of 0.
A more realistic value of c is 2.333, which implies a clean, porous
sandstone.
Even better, a value for c should be estimated from the local
geology. Dillon et al. (TLE, October, 2003) found values as high as
2.8 for sediments in offshore Brazil.
Jan 2008

80

AVO Inversion
Simultaneous Inversion

Aki-Richards Equation
We start with Fattis version of the Aki-Richards equation. This models
reflection amplitude as a function of incident angle:

RPP ( ) = c1RP + c2 RS + c3 RD
where:

c1 = 1 + tan 2
c2 = 8 2 sin 2
1
c3 = tan 2 + 2 2 sin 2
2
V
= S
VP

Jan 2008

RP =

1 VP
+

2 VP

1 VS
+

2 VS

RD =
.
RS =

82

RS and Density Components


A problem with this equation is that the coefficients are not equal in size.
This makes the solution for RS and Density unstable at small angles:

= 30
c1 = 1 + tan 2
c2 = 8 2 sin 2
1
c3 = tan 2 + 2 2 sin 2
2
VS
=
= 0.5
VP
Jan 2008

= 60

1.330
4.000
0.500 1.500

0.041 1.125
Conclusion: the direct solution can be
unstable.

83

Transforming Variables
We want to use the fact that the basic variables, ZP, ZS, and are related.
In simultaneous inversion, we assume this linear model for the background
trend:

ln( Z S ) = k ln( Z P ) + k c + LS

ln( ) = m ln( Z P ) + mc + LD

Ln()

Ln(Zs)
LD

LS

Ln(Zp)
Jan 2008

Ln(Zp)
84

The New Equation


This changes Fattis equation to:

T ( ) = c%1W ( ) DLP + c%2W ( ) DLS + c3W ( ) DLD


where:

c%1 = (1 2) c1 + (1 2) kc2 + mc3


c%2 = (1 2) c2

W ( ) = wavelet at angle
D = Derivative operator
LP = ln ( Z P )

Jan 2008

85

Whats the difference?


The new equation is better than the old because:
(1) We now have independent variables, which makes the system more
stable.
(2) We have built in the known regional rock property relationships
between variables for the background case.
(3) We can apply independent pre-whitening or stabilization to the
variables LS and LD to handle noisy traces.

Jan 2008

86

Simultaneous Inversion Theory


The algorithm looks like this:
(1) Given the following information:
- A set of N angle traces.
- A set of N wavelets, one for each angle.
- Initial model values for Zp, Zs, and .
(2) Calculate optimal values for k and m using the actual input logs.
(3) Set up the initial guess:

[ LP

LS

LD ] = [log( Z P ) 0 0]
T

(4) Solve the system of equations by conjugate gradients.


(5) Calculate the final values of Zp, Zs, and :

Z P = exp( LP )
Z S = exp(kLP + kc + LS )

= exp(mLP + mc + LD )
Jan 2008

87

Synthetic Test

Vp

Vs

Synthetic

As a test, we produced
100%
a series of synthetic
Gas
gathers corresponding
to varying fluid effects.
Target Zone
The synthetics were
created using BiotGassmann substitution
and elastic wave
modeling.
Two of these synthetics
are shown here.

100%
Wet

Target Zone

Jan 2008

88

Result at the Gas Location


Initial guess:
Zp

Zs

Model

Input

Error

After 50 iterations:
Zp

Jan 2008

Zs

89

Result at the Wet Location


Initial guess:
Zp

Zs

Model

Input

Error

After 50 iterations:
Zp

Jan 2008

Zs

90

Real Data Test - Colony


This test applies the simultaneous inversion algorithm to the Colony data
set from Western Canada:

Jan 2008

91

Transform to angle gathers and read into STRATA:

Jan 2008

92

Using the known well, create cross plots to determine the optimum
coefficients:

Jan 2008

93

Real Data Test Inversion Results


Zp

Zs

Jan 2008

94

Zp

Vp/Vs

Jan 2008

95

Real Data Test Error Analysis


Input gathers:

Synthetic data from inversion:

Jan 2008

96

Input gathers:

Synthetic error from inversion:

Jan 2008

97

Comparison
between real logs
and inversion
result at well
location

Jan 2008

Zp

Vp/Vs

98

Cross plotting Vp/Vs against Zp using the log curves:

This zone should


correspond to gas:

Jan 2008

99

Zp

Vp/Vs

Gas Zone
from log
cross plot
Jan 2008

100

The Colony Gas Sand


Simultaneous Inversion

In this exercise, we will start with the 2D Colony dataset which we analyzed earlier
and apply the Simultaneous Inversion analysis to it.
For this exercise, we will use the STRATA program to invert the angle gathers. If the
AVO program is still running, close it down by clicking on File / Exit Project on any
of the AVO windows.
On the GEOVIEW main window, click on the
STRATA button to start the STRATA program:

Select the option to Open Previous Project and


choose the colony.prj project:

Jan 2008

102

The STRATA window now appears,


initially blank.
The first step is to read in the
angle_gathers, which are already stored in
the project. Click on Data Manager /
Import Data / Open Seismic / Open From
Project:

From the list, select angle_gather.vol


and click on Open:

Jan 2008

103

The angle gather now appears, with the single horizon previously picked.

Jan 2008

104

To build the initial model for STRATA, we need more horizons, so we will import
them from a file.

First, delete the existing horizon by


clicking on Horizon / Delete Horizons.

From the list, select Horizon 1, and


click on Ok:

Jan 2008

105

Then, click on Horizon / Import


Horizons / From File.

Select the file


angle_gather_horizons.txt
and click on Ok.

Jan 2008

106

This file has multiple horizons, so click


that option and then click Next>>.

There are 3 horizons, so fill in that


number, and click Next>>

Finally, fill in the format page as


shown to the right. You may want
to Display selected file to verify
these choices. When you have
completed the menu, click on Ok to
load the horizons.

Jan 2008

107

Now, we build the initial model for the


inversion. Click on Model / Build /
Rebuild a Model:

On the first page choose the option


Typical setup for Pre-stack Inversion
and click Next>>:

On the next page, we confirm the


wells used in the model. Click on
Next>>.

On the next page, we confirm


which logs curves are used.
Click on Next>>.
Jan 2008

108

On the next page, we confirm which


horizons are used in the model
building. Click on Next>>:
Finally, we confirm that the
model will be filtered to retain
only the low frequency
components. Click on Ok.

When the model has been


built, it will look like this:

Jan 2008

109

Because this is pre-stack seismic data, the default


display does not show the model in a continuous
form. To verify the model, click on the eyeball icon:

On the View Parameters menu, turn off the


trace plotting (temporarily) by changing the
Trace Data Volume to None and clicking Apply:

The STRATA window now shows


the low frequency impedance
model which will be used for the
inversion.
Click on Cancel on the View
Parameters menu to restore the
original STRATA window.
Jan 2008

110

We are now ready to do the simultaneous inversion. This is


actually done in two stages. First we apply inversion at the
well location(s) to confirm the inversion parameters and
allow the program to determine the proper scaling. Then, we
apply inversion to the entire volume. To do the first step,
click on Analysis / Pre-stack Analysis:

On the first menu page, select


angle_gather as the input and click
Next>>:

On the second menu page, we confirm


the angle range for this data set. Click
Next>>:

Jan 2008

111

On the next page, we confirm a


number of parameters. The most
important parameter on this page
is the wavelet. By default, STRATA
will use the last extracted wavelet.
To display that wavelet, click on
Set Current Wavelet:

When the wavelet menu appears,


it displays that last wavelet we
extracted, wave2. Click on Cancel
to accept that wavelet and remove
this menu.

Now click on Next and Ok on this


group of menu pages.
Jan 2008

112

Now, a new menu appears, allowing you to set the background relationship
between ln(ZP), ln(ZS), and ln(Density):
These cross plots have
been calculated using
the full range of the logs
from the AVO well. An
improved estimate could
be made by limiting the
depth range of the data
being used.
For now, we will simply
manually improve the
regression fit through
the clusters.

Jan 2008

113

Using the mouse,


modify the
regression lines
from this:

To this:
And click Ok and
Save regression
coefficients:

Jan 2008

114

The Pre-stack Inversion


Menu now looks like this
(note that the coefficients
may not be exactly the
same):
Most of the parameters on
this menu can be defaulted.
However, because the
maximum angle range is only
30 degrees, it is probably
advisable NOT to solve for
density. For that reason
toggle OFF Update Density:
Then click on Apply to see the
inversion result at the well:

Jan 2008

115

The display shows a number of useful curves:

Real Log

Initial
Model

Inversion
Trace
Jan 2008

Synthetic
Error
Real
Data

116

To customize the display, click on the eyeball icon:

On the Layout page, remove


the plot of Density and add
the plot of Vp/Vs:

Then, click on the Curves page:

And select the option to


Apply a filter to the
original logs for display.
Finally, click on Ok.
Jan 2008

117

The new display shows a very good fit between the inversion traces and the original
logs, especially near the target zone:

Jan 2008

118

Now we will apply the inversion to the entire data set.


Click on Inversion / Pre-stack Inversion:

The menu that follows confirms all


the parameters we have already
seen. So we can default every
page, except the last one, which
determines which volumes will be
created. Since we have chosen
NOT to update Density, we remove
it from the list and add Zs instead
Now, click Ok to create the
inversion volumes:

Jan 2008

119

A series of windows now appears.


One window is the synthetic data
corresponding to the inversion
output.
We can also see the error, which
is the difference between the real
data and synthetic data.
Click on the eyeball and set the
Trace Data Volume as shown
below:

Jan 2008

120

The other volumes


are Zp, Zs, and
Vp/Vs:

Jan 2008

121

A very helpful analysis can be done


by cross plotting Zp against Vp/Vs
from the well logs. To do that, click
on Well / Edit / Correlate Well:

Select the well and click Edit:

Jan 2008

122

When the eLog window appears,


click on Crossplot / Vp/Vs vs AI :

Accept the defaults on


most of the menus which
appear:

Jan 2008

123

On the final menu page, set


the depth range around the
zone of interest as shown:

Jan 2008

124

The cross plot shows a very


strong separation between the
background wet trend and the
hydrocarbon zone. To
highlight that zone, click on
Zones / Add:

Jan 2008

125

Under the Current Filter


Selection, select Add new filter:
Change the name of the filter as
shown:
We will only define one zone, so
change the starting color:

Select red for this zone:

Jan 2008

126

Now highlight a region around the


anomalous points, and click Apply and Ok.
Click on the Cross-section button to
confirm that the highlighted zone
corresponds to the gas sand:

Jan 2008

127

Now go to the window containing


the Zp inversion result. To use
the cross plot, we need to open
the Vp/Vs volume into the same
window. To do that, click on File
/ Open Seismic / From Project:

From the list select


inverted_VpVs.vol and click Open:

Jan 2008

128

The window now contains


both volumes, as shown in
the label. Click on the
eyeball icon:

Change the View


Parameters menu as
shown and click on Ok:

Jan 2008

129

The resulting plot now shows the gas sand region highlighted, as expected. (If your
display does not show the red anomaly, it may be covered by the displayed horizon. In
that case, use the eyeball to remove the displayed horizons temporarily.)

Jan 2008

130

A second useful analysis is LMR (lambda-mu-rho)


analysis. To do that, go to the window containing the
Zp and VpVs volumes and click on Process / LMR
Transform:

On the resulting menu, select the proper volumes for


the analysis. All other parameters can be defaulted:

Jan 2008

131

The calculation produces both lambda-rho (lmr_LR) and mu-rho (lmr_MR) in the same
window:
Mu-rho
By manipulating the
plot settings, we can
see the individual
results:

Lambda-rho

Jan 2008

132

The best way to analyze the LMR


volumes is to cross plot them:

On the cross plot menus, set the


parameters shown:

Jan 2008

133

According to the LMR theory, the


hydrocarbon region should
correspond to the low lambdarho values.
As before, we will add a single
zone in that region.
Click on Zones / Add:

Jan 2008

134

On the menu, Add a new Filter


Selection:

Rename the Filter Selection


LMR Zone:

Using the mouse highlight a


rectangular region, something like this
and click Ok on the menu:

Jan 2008

135

Now, go to the window containing


the lambda-rho and mu-rho
volumes and click on the eyeball:

Modify the View Parameters menu


as shown. Note that you may have
to click the Zone filter list button to
make sure you are using the LMR
zone.
When you have completed the
menu as shown click on Ok.

Jan 2008

136

Once again, the selected zone highlights the anomalous gas sand region:

We have now completed the Simultaneous Inversion exercise. Close down the STRATA
program by clicking on File / Exit Project on any of the windows.
Jan 2008

137

Case Study
Simultaneous Inversion
on the Marlin Field

Case Study : Marlin Field


Seismic amplitude map

Discovery well A1
150 feet gas sand

Up dip delineation well A6


80 feet gas sand over 60 feet wet sand

Down dip delineation well A5


140 feet wet sand

Project goal
Predict sand, porosity and fluid

Data used
A5 and A6 wells; time migrated
gathers and velocity

Can inversion distinguish the


wet well from the gas wells?
Jan 2008

139

TVD
FT

VSH_FIN
v/v

PHIE_ND
v/v

PR
v/v

AI-PR CROSSPLOT,
in-situ case

0.5

AI
0

10000

35000

0.8

11000

35000

( COREPHI )
0
100

0.6

25000

Shale

Vshale

ZP
0.4

Shaley Sand, wet


15000

11200

ACOUSTIC IMPEDANCE

11100

Gas Sand
0.2

11300

0.1

0.2

5000

11400
0.3

0.4

0.5

POISSON'S RATIO

11500

Jan 2008

By cross plotting ZP against for 2 wells, we verify that we


should be able to distinguish gas from wet sands and shales.140

Synthetic Gathers with Amplitude


Plot

INSITU

BRINE

OIL

GAS

TOP
BASE

BASE

TOP

Synthetics confirm that we should be able to distinguish gas from wet sands and
shales.

Jan 2008

141

AVO Analysis: Data Enhancement


Band pass filtering
Trim static correction
Radon noise reduction
AVO background trend
correction
Super gather

BEFORE

angle

Jan 2008

angle

45

BEFORE

AFTER

45

AFTER

BEFORE

AFTER

It is important to correct the input seismic


data for noise, time alignment, and
background amplitude.

angle

45

BEFORE

angle

AFTER

45

142

AVO Analysis: AxB and Fluid Factor


AVO AxB

AVO FF
?

!!

!!

Neither of these conventional AVO attributes were unambiguous


in differentiating gas from wet sands.
Jan 2008

143

Pre-Stack Inversion: Inverted


P-Impedance
Seismic amplitude map

Jan 2008

Inverted P impedance map

144

Pre-Stack Inversion: Inverted SImpedance


Seismic amplitude map

Jan 2008

Inverted S impedance map

145

Pre-Stack Inversion: Inverted


Density
Seismic amplitude map

Jan 2008

Inverted density map

146

Comparisons at Used Well Locations

P impedance

P impedance

S impedance

S impedance

Density

Density

Jan 2008

A5 well

A6 well

147

Prediction Against Blind Test Well A1


P imp log
P imp Inv

S imp log
S imp Inv

Den log
Den Inv

0
0

Jan 2008

Lith Inv
VSH log

1 0

PHIE Inv

50 0

SW Inv

PHIE log

50

SW log

These are calculated Lithology, Porosity and Water Saturation


curves at the blind well compared to known values.

148

Sand Prediction
By plotting against Vshale
from the known logs, we
conclude that the lower left
quadrant is clean sand:

Vshale
This allows us to calculate the
Sand % volume from the
Inversion Density volume.
Jan 2008

149

Porosity Prediction
By plotting against ZP from
the known logs, we derive 2
porosity trends:

Wet trend

Pay trend

ZP
This allows us to calculate the
Porosity volume from the
Inversion P-impedance volume.
Jan 2008

150

Water Saturation Prediction


Using the Inversion Density
volume and the Derived
Porosity volume, we calculate
Water Saturation using:

Wet well

= ( (1 SW ) Gas + SW Water )
+ (1 ) Matrix
where:

Gas = 0.1
Water = 1
Matrix = 2.65

Jan 2008

Gas well

The Derived Water Saturation volume correctly distinguishes the


wet well from the gas wells.

151

Conclusions
(1) Simultaneous inversion of pre-stack gathers can be used to predict
P-impedance, S-impedance and Density volumes.
(2) The original Aki-Richards equation can be very unstable for inversion.
(3) By transforming variables to include regional rock property trends, we
have shown how to stabilize the system.
(4) We have applied the algorithm to a model data set, which successfully
discriminated between wet and gas cases.
(5) We have shown the successful application to a case study from the
Gulf of Mexico.

Jan 2008

152

Processing Issues in AVO

Introduction
The processing of Seismic Data can have a critical impact on the AVO
analysis. In this section, we address the following issues, as they relate to
AVO:
Trace Mute.
Noise Attenuation:
Super Gather.
Multiple Suppression.
Random Noise Suppression.
Scaling:
Surface Consistent Scaling.
Model Based Scaling.
Residual Moveout:
Higher Order Moveout.
Time-Variant Trim Statics.
Pre-stack Migration.
Jan 2008

154

Trace Mute
The purpose of a Trace Mute in processing is to limit the influence of long
offset traces, where the noise level is high.

Raw Gathers

Jan 2008

After Mute

155

Trace Mute
In traditional processing, the mute deals with 3 types of far-trace noise:

NMO stretch

High amplitude
noise bursts

Residual
Moveout

The goal of conventional processing is to produce an


optimum CDP stack.
This usually requires a tighter mute than is desirable for
AVO analysis.

Jan 2008

Recommendation: start with un-muted gathers, and mute in


the AVO program.

156

Noise Attenuation
Noise can be divided into two broad categories.
The first type is random noise,
which is uncorrelated spatially, and
cannot be easily modeled:

The second type is coherent noise,


typically caused by multiples:

In AVO, the challenge is to suppress the noise, while preserving the


amplitudes of the signals.
Jan 2008

157

Super Gather
The Super Gather is a
very robust tool for
reducing random noise.

Input gathers

It is very good at
preserving offsetdependent amplitude
variations.
However, it can smooth
out structural variations.

After Super Gather

Jan 2008

158

Super Gather
The Super Gather has 2 important
parameters.
The Number of Offsets determines how
many offsets will appear in each final
gather. Normally, this is set equal to the
average fold of the input data.

Xline

The Rolling Window determines which


adjacent input CDPs will be averaged to
produce an output CDP. This determines
the degree of random noise suppression,
and also, the degree of lateral smoothing.

Jan 2008

Inline

159

INVEST Parabolic Radon Transform


INVEST is a noise suppression tool, based
on the Parabolic Radon Transform.
INVEST can be used for random noise
suppression or coherent noise
suppression or both.
INVEST assumes that, after NMO (or prestack migration), all coherent events can
be modeled as parabolic shapes.

Primary

Multiple

At each time sample, INVEST uses a fan


of possible parabolas to model the events
actually found in the data. The user
specifies the range of this fan.

Jan 2008

160

INVEST Parabolic Radon Transform


The INVEST fan is divided into two
components based on the users cutoff
specification. All events with a moveout greater than the specified value are
assumed to be multiples.

Primary
Cutoff
Multiple

Actually, the parabolas do not


completely model the data.
So, everything that is not modeled by
parabolas is assumed to be random
noise:

Seismic Gather = Modeled Primary + Modeled Multiple + Random Noise

Jan 2008

161

INVEST Parabolic Radon Transform


An individual parabola is defined by its
move-out at the far offset, which is
expressed in ms:

Delta-T in ms

This menu is used to define the range of parabolas and cutoff:

Offset
Primary

Multiple

Jan 2008

162

INVEST Parabolic Radon Transform


Once the model is defined, there are two options for noise suppression:
1. Multiple elimination:
Solve for Modeled Primary + Modeled Multiple.
Subtract the Modeled Multiple from the original gathers.

Input
Jan 2008

After Multiple Suppression

Removed Multiples
163

INVEST Parabolic Radon Transform


The second noise suppression option is:
2. Random Noise Suppression:
Solve for Modeled Primary + Modeled Multiple.
Subtract both Modeled Primary + Modeled Multiple from the original gathers. This
gives the Random Noise.
Scale the Random Noise to the desired level.
Subtract the Scaled Noise from the original gathers.

Input
Jan 2008

After Noise
Suppression

Removed
Noise
164

INVEST Parabolic Radon Transform

Summary of Parabolic Radon Transform:


The Radon Transform can remove either multiples or random noise or
both.
In either mode, the primary amplitudes can be distorted, if parameters
are not chosen carefully. Testing is critical.
In the noise suppression mode, the user has greater control over the
final amplitude balance.

Jan 2008

165

Surface Consistent Scaling

Surface Consistent Scaling assumes that the scaling correction for each
trace can be written as the sum of a source component and a receiver
component:
ScalerTrace = ScalerSource + ScalerReceiver
The scaling coefficients are determined by averaging in the common receiver
and common shot domains, rather than computing a single scalar for each
trace.
The effect of averaging is to remove lithologic or offset-dependent variations,
which would distort the AVO effects.
The next four slides, from Downton (2005), compare the effects of surface
consistent scaling versus single trace mean scaling.

Jan 2008

166

Time (s)

CDP Gathers (mean scaling)

Ostrander Gathers (Relative Amplitude Processing)


Jan 2008

Downton and McKidd, 1997

167

Time (s)

CDP Gathers (surface consistent scaling)

Ostrander Gathers (True Amplitude Processing)


Jan 2008

Downton and McKidd, 1997

168

S-Reflectivity (mean scaling)

Time (s)

Jan 2008

Downton and McKidd, 1997

169

S-Reflectivity (surface consistent scaling)

Time (s)

Jan 2008

Downton and McKidd, 1997

170

Residual NMO (RNMO)

If there is RNMO present, this will distort the estimate of the gradient and
any other related attributes. RNMO may arise from:
Sparse nature of velocity analysis
Azimuthal variations
Ignoring higher order terms in NMO correction
Ways of reducing RNMO:
Denser velocity analysis
Higher order corrections
Automatic velocity analysis
In the next sequence of slides, the effect of RNMO on the intercept and
gradient will be illustrated (Ratcliffe and Roberts, 2003).

Jan 2008

171

Calculation of Intercept and Gradient

offset

Pre-stack CMP Gather

Amp.
Intercept

Gradie
n
Jan 2008

offset
172

AVO Attributes using Correct Velocity

offset

Correct Velocity

Intercept

Gradient
Ratcliffe and
Roberts, 2003

Jan 2008

173

AVO Attributes using Wrong Velocity

offset

Incorrect Velocity

Intercept

Gradient
Ratcliffe and
Roberts, 2003

Jan 2008

174

Higher Order Move-out


One of the causes of RNMO in conventional processing comes from ignoring
higher order move-out terms:

t X2 = t02 + A2 x 2 +
Conventional
Dix NMO

A4 x

A2 =

1 +
V0 t 0

where:

1 2
,
V02

A4 =

2( )
.
t 02V04

, = Thomsen's
Higher Order
NMO

parameters.

Tsvankin and Thomsen (1994) showed that the 4th order term may be related
to anisotropy in a VTI medium.
Jan 2008

175

NMO Comparison (to 45 )


NMO Curves
Offse t
0

500

1000

1500

2000

-0.800
-0.850
-0.950
Time (sec)

A comparison between the 2term and 3-term curves:

-0.900
-1.000
-1.050
-1.100

Higher order curve

-1.150
-1.200
-1.250
-1.300
NMO

Dix 2-term curve

NMO/TIV

The problem comes when the


2-term curve is used to correct
the 3-term curve:

Time (msec)

NMO/TIV Difference
60
50
40
30
20
10
0
-10 0
-20
-30

NMO-corrected
higher order
curve
500

1000

1500

2000

Offset (m) ( Far = 45 degrees)


NMO/TIV - NMO

Jan 2008

176

NMO Comparison on real data

2-term NMO
applied to real
gathers with
long offset:

3-term NMO
applied to real
gathers with
long offset:

Jan 2008

C.P. Ross, 1997

177

Time-variant Trim Statics


Conventional Trim Statics is the process of calculating a single time
shift for each trace in the gather to align traces within a time window:

Analysis
window

Jan 2008

Stack to
produce a
single pilot
trace.

Calculate
time shifts
by cross
correlation
and apply.

178

Time-variant Trim Statics


Conventional Trim Statics is limited to a single window, which makes it difficult to align
separated events. Time-variant Time Statics performs the calculation over a series of
smaller, overlapping windows, and interpolates the calculated shifts:

Analysis
windows
Stack to
produce a
single pilot
trace.

Jan 2008

Calculate
time shifts
by cross
correlation.
Interpolate
between
windows.

179

Time-variant Trim Statics

Strength:
Very simple and fast to apply.
Does not require knowledge of the previously applied velocity field.
Corrects any type of time mis-alignment.

Weakness:
Because there is no physical model, shifts may be unrealistic.
Some AVO anomalies look like time shifts, but are really amplitude/phase shifts.
Needs to be QCd carefully.

Jan 2008

180

Gulf Coast Exercise


Part 1: Data Preparation

Gulf Coast Exercise 1


The next series of Exercises
uses a dataset from the US Gulf
of Mexico, described at right.

Location: SE of New Orleans LA


Water Depth: 1500 ft
Reservoir Sands: Turbidite channel-levee sequence
Production: Gas and condensate

There are 3 exercises, which


cover a complete AVO project
on this 3-D volume:
(1) Data Preparation
(2) AVO Modeling
(3) AVO Analysis

Jan 2008

182

Open the Well Database


First we will open an existing database. To do that, click Database / Open on
the GEOVIEW window.

Select the database, GOM


Workshop and click OK.

Jan 2008

183

The Well Explorer shows that we have 2 wells loaded within this database. All
wells are deviated, which is indicated by the symbol shown beside the well
name.

Jan 2008

184

View the Base Map


To see the well locations on a
base map, click on the Base Map
tab.

Two of the wells,


Well_1 and Well_1_ST1,
will be contained within
the 3-D survey which
we are about to
analyze.

Jan 2008

185

Use the 3-D Viewer


Find the small cube icon on
the top of the Base Map and
click it.

This will display the wells within a 3-D


visualization window.
This window can be manipulated in a
number of ways.
For example, hold down the left
mouse button and move the mouse to
rotate the display.
Hold down the right mouse button
and move the mouse to move the
display.
Jan 2008

186

The tool bar on the left contains


other tools for manipulating the
display. Some of these are
highlighted below:

Zoom
After examining the 3-D
viewer, click on the x on
the upper right of the
window to dismiss it.

Jan 2008

Unzoom
Turn on
annotation of
well names.

187

Use the Curve View


Another useful display
is the Curve View:

To use that, you must


expand one of the wells
(say, Well_1) by clicking
on the + sign next to the
well name.
Then double-click any
of the log curves to see
that curve displayed.
This shows the sonic
log in Well_1.
Jan 2008

188

Use the Table View


Finally, go back to the Table View to see a detailed list of all the curves in
Well_1. To see a display of all the curves, click on the Display Well button on
the lower right.

Jan 2008

189

Use the Log Display Window


The target zone for
this well is a gas
saturated sand at
about 9750 ft below
sea level, or about
2550 ms on the
correlated log.
The region is
identified by the
two markers, Pay
and Base Pay.
The well does not
contain any shear
wave information,
so this will have to
be calculated.
Jan 2008

190

Start the AVO Project


Now we will start the AVO Project. To do
that, click on AVO / AVO on the Geoview tool
bar:

Select the option to Start New Project and


click on Ok:

Type in the new project


name, GOM Project, and
click Ok:
Jan 2008

191

Load the Seismic Data

On the AVO Analysis Window, start


loading the seismic volume as shown:

Select the file RawGathers and click on


Next>>:

Jan 2008

192

Load the Seismic Data

Default the selection on the next page,


which specifies that we are reading a 3D
file. Click on Next>>:

Accept all the defaults on the


following page. Note that we are
assuming there are both Inline/Xline
numbers as well as X-Y coordinates
in the trace headers. Click on Next>>:

Jan 2008

193

Load the Seismic Data


The SEG-Y Format page needs
to be changed. The default
parameters are the standard
SEG-Y byte locations for various
values, as shown on the right.

However, this file has values in


different locations. Change the
values as shown on the right.
Then click on Next>>.

Jan 2008

194

Load the Seismic Data


Click on Yes on the confirmation
menu to begin scanning the file.

Finally, the Geometry Grid Page


should show that the file has been
read properly.
Check that your geometry page
looks identical to the one on the
right.
If not, click on Back and check that
you have changed the byte locations
correctly.
When your Geometry Grid Page
looks correct, click on Ok to
complete loading the file.
Jan 2008

195

Load the Seismic Data


The Well To Seismic Map Menu
shows that only two wells are
contained within this seismic
grid. They have been positioned
correctly within the grid, so
click on Ok to accept this
mapping.

Jan 2008

196

Load the Seismic Data


When the seismic window appears, click on the - icon to reduce the
plot scale.
Then, type in the Inline 14770 and position the display at the well
location as shown:

Jan 2008

197

Read in a Horizon
The final data loading task is to load a
previously picked horizon. To do that,
click on Horizon / Import Horizons / From
File:

Select the file M4 Sand T and click Ok.

On the next page, accept the defaults,


but click on the button Display the first
selected file. Then click on Next>>.
Jan 2008

198

Read in a Horizon
The displayed file shows that the
format must be specified as
shown below. When you have
made those changes, click on Ok
to read in the horizon file.

Jan 2008

199

Read in a Horizon
This is the new display with the horizon positioned at the zone of interest.

Jan 2008

200

Noise Problems
The data shows two problems, which need to be addressed .

One problem is the noise, especially


on the far traces.

A second problem is that the target


events have not been properly
moveout-corrected.
Jan 2008

201

Apply a Mute
First, we apply a mute to remove most of
the far trace noise. Click on Process /
Utility / Mute.
On the first page, accept all the defaults.
Click on Next>>.

Jan 2008

202

Apply a Mute
On the second page, we will set
parameters for the Outer Mute, ie, a mute
on the far offsets. Click on Open Outer
XT Table:

Fill in the Mute Table


as shown. This
corresponds to the
mute shown on the
right.
Then, click on Ok on
this menu and on the
Mute Process Menu to
apply this mute to the
data.
Jan 2008

203

Apply a Mute
When the Mute process has finished, the result looks like this:

Before Mute

Jan 2008

After Mute

204

Apply a Filter
After the mute, we can still see some residual
low frequency noise, crossing the zone of
interest. We will try a band pass filter to
reduce that noise.
First, examine the amplitude spectrum of the
input data by clicking on Process / Utility /
Amplitude Spectrum:

Jan 2008

205

Apply a Filter
Fill in the menu as shown. This
examines a single CDP gather near
the well:

The displayed amplitude


spectrum shows significant
energy below 5 Hz, which is
probably noise.
We will design a filter to
reduce the low-frequency
noise, while leaving the highfrequency alone.
The indicated filter
corresponds to the setting:
5/10-30/80
Jan 2008

206

Apply a Filter
To apply the filter, go to the
window containing the mute data,
and click on Process / Filter /
Bandpass Filter:

On the first menu page, leave all


the default parameters, except to
change the output file name to
mute_filter. This will easily allow
us to identify the volumes.
Click on Next>>.

Jan 2008

207

Apply a Filter
On the next menu page, type in the
desired filter parameters as
shown, and click Ok.
When the process has completed,
we can see an improvement at the
event of interest, especially at the
far offsets.

Before Filter

Jan 2008

After Filter

208

Super Gather
A very useful random noise
suppression operation is the
Super Gather. To start that, go to
the window containing the
mute_filter data and click on
Process / AVO(Prestack) / Super
Gather:

On the first page, set the Output


file name to mute_filter_super:

On the second page, set the


parameters as shown. Click on Ok
to create the Super Gather.
Jan 2008

209

Super Gather
The Super Gather, in this case, has had a more subtle effect,
but has reduced the random noise somewhat:

Before Super Gather

Jan 2008

After Super Gather

210

Trim Statics
The final processing step we will perform is
Trim Statics. The purpose of this step is to
correct the residual time alignment problems,
especially at the zone of interest.
Go to the window containing the
mute_filter_super, and click on Process /
AVO(Prestack) / Trim Statics:

On the first page, keep all


the default parameters.
Note that we are creating an
output file named
mute_filter_super_shifted.

Jan 2008

211

Trim Statics
On the second page, fill in the
parameters as shown. Note that
we are using a series of 100 ms
windows, with a 50 ms overlap.
Also, the maximum allowed static
shift is 50 ms

When the process finishes, we see


that it has done a very good job of
aligning the data:

Before Trim Static

Jan 2008

After Trim Static

212

Clean up Display
We have now completed the preliminary processing of the AVO data. Of the
various windows created, we only need to display the final,
mute_filter_super_shifted.
Close all other seismic windows, by clicking on the x at the upper right
hand corner:

Jan 2008

213

Gulf Coast Exercise


Part 2: AVO Modeling

Start AVO Modeling Window


Now we are ready to do AVO
Modeling with the well which ties our
anomaly. To start that, click on
Modeling / Single Well.

Select Well 1 from the list and click


Open.

Jan 2008

215

Start AVO Modeling Window

Accept the default P-wave and


Density logs. Click Ok on this menu.

As before, the program warns us that


there is no S-wave log available, so
Castagnas equation will be used to
create it. Click Ok on this menu.
Jan 2008

216

Start AVO Modeling Window


On the Transforms Menu, accept all
the defaults, which will use
Castagnas equation with the textbook coefficients. Click Next and Ok.
When the AVO Modeling window
appears, we see that the derived
Poissons Ratio curve does not show
the expected behavior in the Pay
Zone. This is because Castagnas
equation does not apply in the zone.
We need to use Fluid Replacement
Modeling to correct that zone.

Jan 2008

217

Perform FRM
To start the FRM process, click on the
FRM button.

The input page should be filled as


shown to the right.
Note that we are specifying a gas
reservoir with 30% brine
saturation.
Click on Next.

Jan 2008

218

Perform FRM
We will default all parameters on the
subsequent menus. On the second
page, note that we are applying the
analysis to the region outlined by the
tops called Pay and Base Pay.

On the last page, click on the QC


Display button.
Note that the display shows a
significant drop in Poissons Ratio
within the zone of interest.

Jan 2008

219

Perform FRM
Finally, click the Ok button and all the
menus as shown. This will apply the
FRM calculation to create the new
logs.

Jan 2008

220

Perform FRM
As expected, the FRM calculation has changed the S-wave and Poissons Ratio
logs, within the target zone:

Before FRM

After FRM

Jan 2008

221

Insert the Seismic Volume


To insert the seismic gathers into the Modeling
Window, click on Seismic / Open From Project:

Select the seismic volume,


mute_filter_super_shifted, and click on
Open to see the seismic inserted.
One problem we can see is that the scale
is not optimal:

Jan 2008

222

Insert the Seismic Volume


To improve the display, click the eyeball icon once again:
Make sure the Seismic Views tab is showing and that the
volume mute_filter_super_shifted is selected:

Then change the Trace


Spacing, as shown and click
Ok:

Jan 2008

223

Insert the Seismic Volume


The display will now look like this:

Jan 2008

224

Log Correlation
Now, we will correlate the well and extract a
wavelet. To start that, click on Logs / Correlate:

The program will extract a


composite trace around the
borehole.
Make sure that the volume
selected is
mute_filter_super_shifted.
Also, set the maximum offset to
2000 m for this analysis. This
means we are using the near
offsets to avoid any AVO effect at
this stage. Click Ok.
Jan 2008

225

Log Correlation
The Log Correlation window
now shows two new traces,
the zero-offset synthetic trace
in blue and the composite
trace in red:

The synthetic trace will be


greatly improved if we extract
a wavelet from the data.
To do that, click on Extract
Wavelet / Use Well:

Jan 2008

226

Log Correlation
Select Well_1 for the wavelet
extraction:

Set the Offset Range to use


the near traces:

Set the Time Window:

Jan 2008

227

Log Correlation

Accept all the defaults on the


last page and click Ok to
extract the wavelet.

The extracted wavelet will look


like this:

Jan 2008

228

Log Correlation
The synthetic now shows a
good correlation with the
composite trace, with a slight
shift.
To optimize that, click on the
Parameters button.

Set the Cross Correlation


window as shown and click on
Apply.
Note that the program
suggests a -7 ms shift to
optimize the correlation.

Jan 2008

229

Log Correlation
Click on the Apply Shift button
to apply that shift.

The correlation now looks


very good, with no further
stretching required.

Click on the Ok button to


accept this change to the well.

Finally, accept the new name


for the shifted sonic log.
Jan 2008

230

Zoeppritz Modeling
Now we will create 2 synthetics. First use the
Zoeppritz modeling algorithm by clicking on
Synthetic / Zoeppritz.

On the menu, default all


parameters except those
shown. Note that we are
setting the Target Zone
around the zone of interest.
Also, we are restricted the
offset range to
approximately that
spanned by the unmuted
traces on the real gather.

Jan 2008

231

Zoeppritz Modeling
When the Zoeppritz Model has completed, it looks like this. Note that the
modeling predicts a strong AVO anomaly. (Note that your synthetic may
not be scaled optimally, but we will fix that later.)

The second synthetic we wish to make is for the wet case i.e., we need to
modify the logs to show a pure brine behavior and create a synthetic from
that.
Jan 2008

232

Creating the Wet Well


The easiest way to have both gas and wet cases
available is to create a totally new well for the wet
case. Click on Wells / Copy Well.

Give this new well the name Well_1_wet.

Now, a completely new AVO Modeling


Window appears, identical to the first. We
do not have the wet logs yet. To do that,
click on the FRM button on the new window.
Jan 2008

233

Creating the Wet Well


On the FRM menus, keep all parameters the same as before except to
change the output Water Saturation (on the last page) to 100%. Click Ok on
this and all the subsequent menus to apply this change.

The new wet well logs are now modified. In particular, the change in
Poissons Ratio is not as great:

Jan 2008

234

Creating the Wet Well


Changing the logs within the wet well did not
automatically change the synthetic. To create the new
wet synthetic, go to the window showing the wet well
logs, and click on Synthetic / Zoeppritz.
Keep all the modeling parameters identical and create the
new synthetic.

When the wet synthetic


appears, change the
display parameters by
clicking on the eyeball,
and setting the
parameters for each of
the synthetics as shown:

Jan 2008

235

Creating the Wet Well


The final display shows a very big difference between the gas and wet
synthetics. The gas synthetic looks much more like the real data.

Gas Synthetic

Jan 2008

Wet Synthetic

Real Data

236

Creating a Cross Plot


The last modeling step is to Crossplot VP/VS vs
Acoustic Impedance. Click on the item as shown
to the right.

On the Crossplot
menus, default all
the parameters
except the ones
shown.
Use only Well_1,
and set the Time
range to start at
2400 ms

Jan 2008

237

Creating a Cross Plot


The resulting Cross Plot shows a clear separation
into two zones. We will highlight the pay zone to
the lower left.

Click on Zones / Add.

On the menu change


the zone name and
color, as shown.

Jan 2008

238

Creating a Cross Plot


With the menus still showing, use the
mouse to highlight the Pay Zone region,
as shown.

Now click on Apply and Ok to save this


zone.

Click on the Cross-Section button to see


the zone applied to the logs. As
expected, the highlighted region defines
the pay zone on the logs.
Finally, click on File / Exit on the Cross
Plot window. When you see this
message, click on Yes.
Jan 2008

239

Closing the Modeling Windows


We have now completed the modeling section of the GOM project. Close down the
two AVO Modeling Windows by clicking on the x on each window:

Jan 2008

240

Gulf Coast Exercise


Part 3: AVO Analysis

Display Angle Range


In this section, we will be working with the final processed gathers from Part 1.
This is called mute_filter_super_shifted. Go to the window containing that
data.
Our first step is to see what range of incident angles are present in this data
set. To do that, click on the eyeball icon.

On the View Parameters menu,


change the Color Data Volume to
Incident Angle.

Jan 2008

242

Display Angle Range


A menu now appears, asking for the
velocity field which will be used to
calculate the angles.
Select the option Single Log and then click
on the button Select from Database.
On the new menu which appears, click first on Well_1 and then choose the log Pwave_corr, as shown below. Then click on Ok on this menu

Jan 2008

243

Display Angle Range


The Velocity Setup Menu now looks as
shown on the right. Click Ok on this menu
and the View Parameters Menu to see the
Incident Angle plot:

The plot shows incident angles to about 45


degrees at the zone of interest, so this
should be good angle range for AVO
analysis.

Jan 2008

244

Pre-stack Picking
Now, we will do automatic pre-stack
picking, using the imported post-stack
horizon as a guide. First, turn off the color
display by clicking View / Color Traces
Shown:

Then, click on Horizon / Pick Horizons:

On the Horizon Selection Menu, click on


Pick Existing Horizon. This is because we
wish to re-pick the imported horizon M4
Sand T. Then click Ok.
Jan 2008

245

Pre-stack Picking
Click Yes on this menu to see a map
display:

On the large picking menu at the base of


the window, click on Options / Automatic
Picking:

Fill in the menu as shown to


the left and click Ok.

Jan 2008

246

Pre-stack Picking
When the automatic picking finishes, it looks like this. Click on Ok to
accept that result:

Now we will display the picked amplitudes.


Click on AVO Analysis / Pick Analysis:

Jan 2008

247

Pick Analysis
On the first menu page, click ON the option to
display the Gradient Analysis with the picks:

On the second menu page, select the M4 Sand T


horizon:

Default all the subsequent parameters to


produce this plot:

Jan 2008

248

Pick Analysis
The curve fit to the picks is not very
good. This could be a problem of
data scaling. It could also be that we
need to use 3 terms in the Gradient
analysis.

To recalculate the gradient curves,


click on AVO Analysis / Pick Analysis
once again.

On the last menu page, change the


Type of Analysis to A/B/C and click
Ok.
Jan 2008

249

Pick Analysis
We can see that the curve fits are
much better using 3 terms in the
analysis, so we will use that for the
rest of the calculations.

Turn OFF the Pick Analysis display


by clicking the items shown to the
right.

Jan 2008

250

Attribute Analysis
Now, we will perform AVO Attribute
Analysis. Click on Process /
AVO(Prestack) / AVO Attribute
Volume:

On the subsequent
menus, keep all the
default parameters,
except to specify the Pwave Well Log as the
velocity input and
specify that we want
A/B/C (3-term) analysis:
Jan 2008

251

Attribute Analysis
When the AVO Analysis has completed the result looks like this. To make
this display, click on the
sign to zoom in by a factor 2.

By default, the color


shows the product A*B.
This shows a strong
anomaly at the well
location, as expected
from a class 3 AVO
anomaly.

Jan 2008

252

Attribute Analysis
To see an alternate attribute combination,
click on the eyeball and change the Color
Attribute to Scaled Poissons Ratio
Change:

As with the previous display,


the attribute (A+B) also
shows the expected
behavior, which is a decrease
in Poissons Ratio at the top
of the sand and an increase
at the base:

Jan 2008

253

Attribute Analysis
Another way of looking at the anomalous region
is to create a Data Slice. To do that, click on
Process / Slice / Create Data Slice.

On the first menu page, we can


choose from a range of AVO
Attributes to display. Select the
Product (A*B) and click Next.

Jan 2008

254

Attribute Analysis
The next menu page determines the
time region over which we will extract
the slice. Set the parameters as
shown. Click Next and Ok to create
the slice.

The extracted Data Slice shows a


strong AVO anomaly at the target,
Well_1, as expected. Interestingly, the
anomaly does not extend to the sidetrack well, Well_1_ST1.

Jan 2008

255

Pre-stack Inversion
The final stage of this exercise is to
perform pre-stack AVO Inversion. We
will do that within the Strata module.
First, close down the AVO Program
completely, by following the series of
steps shown:

Jan 2008

256

Pre-stack Inversion
On the Geoview menu bar, click on the
Strata icon to start that program:

Select the option to Open the


Previous Project and click Ok:

Jan 2008

257

Pre-stack Inversion
On the blank Strata window which
appears, load the AVO pre-stack
data, by clicking on Data Manager /
Import Data / Open Seismic / Open
From Project:

Select the volume


mute_filter_super_shifted and
click Open:

Jan 2008

258

Pre-stack Inversion
Pre-stack Inversion actually
operates on Angle Gathers.
To create this volume, click
on Process / Angle Gather:

Accept the defaults on all


the menus. Note that we
are creating an angle range
from 0 to 45 degrees. Click
Next and Ok to create the
Angle Gather.

Jan 2008

259

Pre-stack Inversion
The calculated Angle Gather
appears in a new window as
shown:

Now that we have the Angle


Gather, we no longer need
the original Super Gather.
Close that window by
clicking the x on the
upper right corner:

Jan 2008

260

Pre-stack Inversion
Build the initial model for inversion by clicking
on Model / Build / Rebuild a Model:

On the model building menus,


default all the parameters except the
ones shown in the next slides:

Jan 2008

261

Pre-stack Inversion

Jan 2008

262

Pre-stack Inversion
After the model is built, the Strata window looks like this:

Jan 2008

263

Pre-stack Inversion

Before inverting the full volume, we will


perform Pre-stack Analysis at the well
location:

Select the angle_gather as the input to this


process, and accept all the other default
parameters:

Jan 2008

264

Pre-stack Inversion

The first menu which


appears confirms the
background relationship
between ZP, ZS, and Density.
Confirm that the regression
lines look roughly as
shown, and click the button
Ok and Save regression
coefficients.

Jan 2008

265

Pre-stack Inversion

On the Pre-stack Inversion


Menu, accept all the
defaults. Click on Apply to
invert the single trace at the
well location.

Jan 2008

266

Pre-stack Inversion
On the Inversion Analysis
window, we will optimize
the display by clicking on
the eyeball icon.

On the Layout page, turn off


the Density and turn on the
VP/VS ratio:

Jan 2008

267

Pre-stack Inversion

Then, click on the Curves


tab and turn on the option
to Apply a Bandpass filter
to the original logs.

Jan 2008

268

Pre-stack Inversion

The Inversion Analysis


shows a strong drop
in VP/VS ratio at the
zone of interest.

Jan 2008

269

Pre-stack Inversion
Finally, we will apply the Pre-stack Inversion to a
single line of the angle gather volume.
Go to the window containing that volume and click
on Inversion / Pre-stack Inversion.

Accept all the defaults on the


following menus, except to
limit the inversion to a single
Inline, as shown, to reduce
the run-time:

Jan 2008

270

Pre-stack Inversion
When the inversion finishes, a
number of windows appear. The
most important, for our purpose is
the VP/VS Ratio display.
Note the predicted drop in VP/VS at
the zone of interest.

Jan 2008

271

Pre-stack Inversion
We would like to confirm that these
anomalous VP/VS values fall within
the region predicted by our cross
plotting in the previous exercise.

To confirm that, we first


need to have the
computed ZP and VP/VS
volumes showing in the
same window.
On the current window
showing the VP/VS
volume click the option
shown:
Jan 2008

272

Pre-stack Inversion
From the list, select the volume
inverted_Zp and click Open.

Now, on the window containing


both ZP annd VP/VS volumes, click
the eyeball icon and set the
display parameters as shown to
the right.

Jan 2008

273

Pre-stack Inversion

The display shows that our


previously derived zone does, in
fact, highlight the target
hydrocarbon zone.

To see that cross plot again,


click on Process / Cross Plot /
Cross Plot Archives.

Jan 2008

274

Pre-stack Inversion
Select the only cross plot on the
list and click Show to see the
previously generated cross plot.

Note that it is
possible to edit the
highlighted zone and
observe its effect on
the display.

Jan 2008

275

Pre-stack Inversion
To modify the zone, click on Zones
/ Edit.

The grab the handle at the corner of the


defined zone and move it. When you have
modified it as desired, click Apply.
Finally, click on View / Redraw on the
window containing the inversion results

Jan 2008

276

Pre-stack Inversion
This shows a comparison of 2 selected zones. Note that to decide which is more
geologically meaningful would require a more detailed analysis of the logs.

Jan 2008

277

Completing the Project

We have now completed the project.


Close down all the windows, by clicking
the items as shown:

Jan 2008

278

AVO Fluid Inversion :


Analyzing uncertainty in AVO

Overview
AVO Analysis is now routinely used for exploration and development.

But: all AVO attributes contain a great deal of uncertainty there is a wide
range of lithologies which could account for any AVO response.

In this talk we present a procedure for analyzing and quantifying AVO


uncertainty.

As a result, we will calculate probability maps for hydrocarbon detection.

Jan 2008

280

AVO Uncertainty Analysis:


The Basic Process

G
CALIBRATED:

GRADIENT
INTERCEPT
BURIAL DEPTH
AVO ATTRIBUTE
MAPS
ISOCHRON
MAPS

STOCHASTIC
AVO
MODEL
FLUID
PROBABILITY
MAPS

PBRI
POIL
PGAS

Jan 2008

281

Conventional AVO Modeling:


Creating 2 Pre-Stack Synthetics
IN
INSITU
SITU==OIL
OIL

IO

GO

FRM
FRM==BRINE
BRINE

IB

Jan 2008

GB

282

Monte Carlo Simulation:


Creating Many Synthetics

I-G DENSITY FUNCTIONS


BRINE

OIL

GAS

75

50

25

Jan 2008

283

The Basic Model

Shale

We assume a 3-layer model with


shale enclosing a sand (with
various fluids).

Sand

Shale

Jan 2008

284

The Shales are characterized by:

Vp1, Vs1, 1

P-wave velocity
S-wave velocity
Density

Vp2, Vs2, 2

Jan 2008

285

Vp1, Vs1, 1

Each parameter has a probability


distribution:

Vp2, Vs2, 2

Jan 2008

286

The Sand is characterized by:

Shale

Sand

Shale

Brine Modulus
Brine Density
Gas Modulus
Gas Density
Oil Modulus
Oil Density
Matrix Modulus
Matrix density
Porosity
Shale Volume
Water Saturation
Thickness

Each of these has a probability distribution.


Jan 2008

287

Trend Analysis
Some of the statistical distributions are determined from well log trend
analyses:
5000
4500
4000
3500
3000
2500
2000
1500
1000
500
0
0.4
Jan 2008

0.9

1.4

1.9
DBSB (Km)

2.4

2.9

3.4
288

Determining Distributions at
Selected Locations
Assume a Normal distribution. Get the Mean and Standard Deviation
from the trend curves for each depth:
5000
4500
4000
3500
3000
2500
2000
1500
1000
500
0
0.4
Jan 2008

0.9

1.4

1.9
DBSB (Km)

2.4

2.9

3.4
289

Trend Analysis: Other Distributions


5000

Shale Velocity

4500
4000

3.0

3500

2.8

Sand Density

2.6 3.0 Shale Density


2.8
2500 2.4
40%
2.6
Sand Porosity
2000 2.2
35%
2.4
1500 2.0
30%
2.2
1000 1.8
2.0
25%
500 1.6
1.8
0 1.4
20%
0.41.2 1.6 0.9
1.4
1.9
2.4
2.9
3.4
15%
1.4
DBSB (Km)
1.0
10%
1.2
0.4
0.9
1.4
1.9
2.4
2.9
1.0
5%
DBSB (Km)
0.4
0.9
1.4
1.9
2.4
2.9
0%
DBSB (Km)
0.4
0.9
1.4
1.9
2.4
DBSB (Km)
Jan 2008
3000

3.4
3.4

2.9

3.4
290

Practically, this is how we set up the distributions:


Shale:
Vp
Vs
Density
Sand:
Brine Modulus
Brine Density
Gas Modulus
Gas Density
Oil Modulus
Oil Density
Matrix Modulus
Matrix density
Dry Rock Modulus
Porosity
Shale Volume
Water Saturation
Thickness
Jan 2008

Trend Analysis
Castagnas Relationship with % error
Trend Analysis

Constants for the area

Calculated from sand trend analysis


Trend Analysis
Uniform Distribution from petrophysics
Uniform Distribution from petrophysics
Uniform Distribution
291

Calculating a Single Model Response


From a particular model instance,
calculate two synthetic traces at
different angles.

Note that a wavelet is assumed


known.
0o
45o

Top Shale

Sand

Base Shale

Jan 2008

292

On the synthetic traces, pick the


event corresponding to the top of the
sand layer:

Note that these amplitudes include


interference from the second interface.

0o

45o

Top Shale
P1

P2

Sand

Base Shale

Jan 2008

293

Using these picks, calculate the Intercept and Gradient for this model:
I
G

0o

= P1
= (P2-P1)/sin2(45)

Top Shale

P1

45o

P2

Sand

Base Shale

Jan 2008

294

Using Biot-Gassmann Substitution


Starting from the Brine Sand case, the corresponding Oil and Gas Sand models are
generated using Biot-Gassmann substitution. This creates 3 points on the I-G cross plot:

BRINE

GAS

OIL

KGAS
GAS
G

G
I

Jan 2008

KOIL
OIL
G
I

I
295

Monte-Carlo Analysis
By repeating this process many times, we get a probability distribution for
each of the 3 sand fluids:
G

Jan 2008

Brine
Oil
Gas

296

The Results are Depth Dependent


Because the trends are depth-dependent, so are the predicted distributions:

@ 1000m

Jan 2008

@ 1600m

@ 1200m

@ 1800m

@ 1400m

@ 2000m

297

The Depth-dependence can often be understood using


Rutherford-Williams classification
4

5
3

Impedance

Jan 2008

Sand

Shale

Class 2
Class 3

Burial Depth

Class 1

298

Bayes Theorem
Bayes Theorem is used to calculate the probability that any new (I,G) point
belongs to each of the classes (brine, oil, gas):

~
P F I,G

)=

~
~
p I ,G F * P (F )

p (I , G F k )* P (F k

where:
P(Fk) represent a priori probabilities and Fk is either brine, oil, gas;
p(I,G|Fk) are suitable distribution densities (eg. Gaussian) estimated
from the stochastic simulation output.

Jan 2008

299

How Bayes Theorem works in a


Simple Case:
Assume we have these distributions:
Gas

Oil

OCCURRENCE

Brine

VARIABLE
Jan 2008

300

This is the calculated probability for


(gas, oil, brine).

OCCURRENCE

100%

Jan 2008

50%

VARIABLE

301

When the distributions overlap, the probabilities decrease:

OCCURRENCE

Even if we are right on the Gas peak,


we can only be 60% sure we have gas.

100%

50%

VARIABLE
Jan 2008

302

Showing the Effects of Bayes


Theorem
This is an example simulation
result, assuming that the wet
shale VS and VP are related by
Castagnas equation.

Jan 2008

303

This is an example simulation


result, assuming that the wet
shale VS and VP are related by
Castagnas equation.

This is the result of


assuming 10% noise in the
VS calculation
Jan 2008

304

Note the effect on the


calculated gas probability

1.0

0.5

0.0
Jan 2008

Gas
Probability
By this process, we can investigate the sensitivity of the
probability distributions to individual parameters.
305

Example Probability Calculations

Gas

Jan 2008

Oil

Brine

306

Real Data Calibration


In order to apply Bayes Theorem to (I,G) points from a real seismic data set,
we need to calibrate the real data points.
This means that we need to determine a scaling from the real data amplitudes
to the model amplitudes.
We define two scalers, Sglobal and Sgradient, this way:
Iscaled
Gscaled

= Sglobal *Ireal
= Sglobal * Sgradient * Greal

One way to determine these scalers is by manually fitting multiple known


regions to the model data.
Jan 2008

307

Fitting 6 Known Zones to the Model

6
3

Jan 2008

308

Real Data Example West Africa

This example shows a real project from West Africa, performed by one of
the authors (Cardamone).
There are 7 productive oil wells which produce from a shallow formation.
The seismic data consists of 2 common angle stacks.
The object is to perform Monte Carlo analysis using trends from the
productive wells, calibrate to the known data points, and evaluate potential
drilling locations on a second deeper formation.

Jan 2008

309

One Line from the 3D Volume


Near Angle Stack
0-20 degrees

Far Angle Stack


20-40 degrees

Jan 2008

310

Near Angle Stack


0-20 degrees

Shallow producing zone


Deeper target zone

Far Angle Stack


20-40 degrees

Jan 2008

311

AVO Anomaly
Near Angle Stack
0-20 degrees

Far Angle Stack


20-40 degrees

Jan 2008

312

Amplitude Slices Extracted from


Shallow Producing Zone
Near Angle Stack
0-20 degrees

+189

-3500
Far Angle Stack
20-40 degrees

Jan 2008

313

Trend Analysis : Sand and Shale


Trends
5000

3.00

4500
2.75

VELOCITY

DENSITY

Sand velocity

4000

Sand density

2.50

3500
3000

2.25

2500

2.00

2000

1.75

1500
1000
500

700

900

1100

1300

1500

1700

1.50
500

1900

700

900

1100

1300

1500

1700

1900

4000
3.00

Shale velocity

2.75

3000

2.50

2500

2.25

2000

2.00

1500

1000
500

Jan 2008

Shale density

DENSITY

VELOCITY

3500

1.75

700

900

BURIAL DEPTH (m)


1100

1300

1500

1700

1900

2100

2300

2500

1.50
500

700

BURIAL DEPTH (m)


900

1100

1300

1500

1700

1900

314

Monte Carlo Simulations at 6 Burial


Depths
-1400

-2000

Jan 2008

-1600

-2200

-1800

-2400

315

Near Angle Amplitude Map Showing


Defined Zones

Wet Zone 1
Well 6
Well 3

Well 5

Well 7

Well 1

Well 2
Well 4

Wet Zone 2

Jan 2008

316

Calibration Results at Defined


Locations

Jan 2008

Wet Zone 1

Well 2

Wet Zone 2

Well 5

317

Jan 2008

Well 3

Well 6

Well 4

Well 1

318

Using Bayes Theorem at Producing


Zone: OIL
Near Angle Amplitudes

1.0
.80

Probability of Oil
.60

.30
Jan 2008

319

Using Bayes Theorem at Producing


Zone: GAS
Near Angle Amplitudes

1.0
.80

Probability of Gas
.60

.30
Jan 2008

320

Using Bayes Theorem at Target


Horizon
Near angle amplitudes of second event

1.0

Probability of oil on second event

.80
.60

.30
Jan 2008

321

Verifying Selected Locations at


Target Horizon

Jan 2008

322

Summary
By representing lithologic parameters as probability distributions we can
calculate the range of expected AVO responses.
This allows us to investigate the uncertainty in AVO predictions.
Using Bayes theorem we can produce probability maps for different
potential pore fluids.
But: The results depend critically on calibration between the real and model
data.
And: The calculated probabilities depend on the reliability of all the
underlying probability distributions.
Jan 2008

323

AVO Summary

This course has given you a comprehensive overview of the Amplitude


Variations with Offset, or Amplitude Versus Offset (AVO) method, both
theoretically and practically.

We have looked at the theory and practical implementation of a number of


AVO procedures, including:
Fluid Replacement Modeling (FRM).
AVO modeling.
AVO data analysis, including Intercept/Gradient cross-plotting.
AVO inversion, including Elastic Impedance, Lambda-Mu-Rho (LMR),
and Simultaneous Inversion.
Possible concerns in AVO analysis.

Let us now briefly review these various procedures.

Jan 2008

324

Fluid Replacement Modeling (FRM)


Before looking at AVO, we needed to understand the rock physics of the reservoir,
using FRM. The basic use of FRM is to use the Biot-Gassmann equations 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). Recall that the Batzle-Wang equations can be used to create the fluid.
Generally all three logs are changed within the target zone.

Output Logs

Input Logs

VP
Jan 2008

SW = 50%

VS

VP

SW = 100%

VS
325

The second use of FRM is to use the Biot-Gassmann equations 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

Jan 2008

VP
SW = 50%

VP

SW = 50%

VS
326

AVO Modeling

AVO modeling involves the following steps:


Edit logs and perform fluid replacement modeling, as just discussed.
Extract a seismic wavelet.
Correlate the logs and data using either a check shot correction or
interactive stretch-squeeze with the extracted wavelet.
Choose and apply a modeling option: Zoeppritz equations with
primaries only, Aki-Richards equations with primaries only, or full
elastic wave equation modeling, which includes multiples and
converted waves.

Another important option that was discussed was anisotropic modeling,


which is available in the HR AVO program.

The next slide shows and example of AVO modeling.

Jan 2008

327

Once we have modeled the logs, we can build a synthetic using the AVO
modeling options, as shown below.

FRM Modeled Logs

Jan 2008

Synthetic

Real Data

328

AVO Analysis

AVO analysis involves the analysis of pre-stack gathers to look for AVO
anomalies.
The main AVO techniques that we discussed were:
Intercept-gradient (A/B/C) analysis, in which we fit the equation
R() = A + B sin2 + C sin2*tan2 to our data.
The fluid factor technique, in which we estimate RP and RS reflectivity
and look for deviations away from Castagnas mudrock line.
AVO polarization analysis.
Recall that we classified AVO anomalies using the Rutherford-Williams
classification scheme.
Class 1 = Reservoirs have higher impedance than encasing shale.
Class 2 = Reservoirs have similar impedance to encasing shale.
Class 2 = Reservoirs have lower impedance than encasing shale.
The next slide shows an intercept-gradient crossplot analysis of a class
3 anomaly.

Jan 2008

329

AVO Analysis Intercept vs Gradient

Gradient (B)

Base Sand

Here is an example of
the A vs B crossplot
for the Colony sand.

Wet Trend

Top Sand
Intercept (A)

Jan 2008

330

AVO Inversion

In the final part of the course, we discussed a number of approaches to


AVO inversion, and applied several of these approaches to the Colony
sand example.

These approaches were:.


Simultaneous inversion.
The elastic impedance (EI) approach.
Lambda-Mu-Rho inversion

The next few slides illustrate these methods.

Jan 2008

331

AVO Inversion Elastic Impedance

EI Far

Elastic impedance applied to a


well log example.

EI Near

EI Far

EI Near

Anomalous Sand

Jan 2008

332

Here are the near and


far EI inversions for the
Colony sand. Note the
difference at the
anomalous gas sand,
shown in the ellipses.

Far EI Inversion

Near EI Inversion

Jan 2008

333

On the left is a cross-plot of the near and


far EI inversions for the Colony sand, with
the anomalous gas sand zone
highlighted.
EI Far

Below, the highlighted zone has been put


back on the section, showing the extent
of the gas sand.

EI Near

Gas Sand

Jan 2008

334

AVO Inversion Simultaneous


Inversion
Simultaneous Inversion solves the Fattis equation, subject to background
constraints:

RPP ( ) = c1RP + c2 RS + c3 RD
where:

c1 = 1 + tan 2
c2 = 8 2 sin 2

1
c3 = tan 2 + 2 2 sin 2
2
V
= S
VP

Jan 2008

RP =

1 VP
+
2 VP

1 VS
+

2 VS

RD =
.
RS =

335

AVO Inversion Simultaneous


Inversion
We assume these linear relationships hold for the background trend:

ln( Z S ) = k ln( Z P ) + kc + LS
ln( ) = m ln( Z P ) + mc + LD

Ln()

Ln(Zs)
LD

LS

Ln(Zp)
Jan 2008

Ln(Zp)
336

AVO Inversion Simultaneous


Inversion
This allows the
simultaneous
calculation of ZP,
ZS, Density, and
other derived
attributes:

Jan 2008

337

AVO Inversion - LMR


Lambda-rho ()

The result is shown above


and the section is shown
on the right. Note in the
highlighted areas that
shows a large increase at the
gas sand zone and shows
a large decrease.
Jan 2008

From the ZP and ZS inversions,


we can then create the and
sections from the formulas:
= ZS2
= ZP2 2ZS2

Mu-rho ()

338

AVO Inversion - LMR


The versus cross plot is
shown on the left, with the
picked gas sand in yellow.
Mu-rho ()

The section below shows the


position of the gas sand.

Lambda-rho ()

Gas Sand
Jan 2008

339

Summary Flowchart
Input Raw Gathers

Optimum Processing

Log based modeling

Wave
Equation
synthetics

Primaries
only
synthetics

Recon Methods

Partial
Stacks

Gradient/
Intercept

Simultaneous
Inversion
LMR

Crossplotting
Jan 2008

340

References
Aki, K., and Richards, P.G., 1980, Quantitative seismology: Theory and methods:
W.H. Freeman and Co.
Batzle, M., and Wang, Z., 1992, Seismic properties of fluids: Geophysics, 57,
1396-1408.
Biot, M. A., 1941, General theory of three-dimensional consolidation:
Journal of
Applied Physics, 12, 155-164.
Blangy, J. P., 1994, AVO in transversely isotropic media - an overview:
Geophysics, 59, 775-781.
Castagna, J.P., Swan, H.W., and Foster, D.J., 1998, Framework for AVO
gradient and intercept interpretation: Geophysics, 63, 948-956.
Castagna, J. P. and Swan, H. W., 1997, Principles of AVO crossplotting: The
Leading Edge, 16, no. 04, 337-342.
Castagna, J.P., and Backus, M.M., Eds., 1993, Offset-dependent reflectivityTheory and practice of AVO analysis, SEG, Tulsa.
Castagna, J.P., Batzle, M.L., and Eastwood, R.L., 1985, Relationship between
compressional and shear-wave velocities in clastic silicate
rocks: Geophysics,
50, 551-570.
Connolly, P., 1999, Elastic impedance: The Leading Edge, 18, no. 4, 438-452.
Downton, J., and McKidd, D, 1997, Methods to improve AVO extractions:
Jan 2008
341
2004 CSEG Convention Expanded Abstracts, 144-145.

References (cont)
Dvorkin, J., Moos, D., Packwood, J. L. and Nur, A. M., 1999, Identifying patchy
saturation from well logs: Geophysics, 64, 1756-1759.
Fatti, J. L., Vail, P. J., Smith, G. C., Strauss, P. J. and Levitt, P. R., 1994,
Detection of gas in sandstone reservoirs using AVO analysis: A 3-D seismic
case history using the geostack technique: Geophysics, 59,
1362-1376.
Foster, D., Smith, S. W., Dey-Sarkar, S. K. and Swan, H. W., 1993, A closer
look at hydrocarbon indicators, 63rd Ann. Int. Mtg: SEG, 731-733.
Gardner, G.H.F., Gardner, L.W., and Gregory, A.R., 1974, Formation
velocity
and density - the diagnostic basis for stratigraphic traps:
Geophysics, 39,
770-780.
Gassmann, F., 1951, Elastic waves through a packing of spheres:
Geophysics, 16, 673-685.
Goodway, B., Chen, T. and Downton, J., 1997, Improved AVO fluid detection
and lithology discrimination using Lam petrophysical parameters, 67th
Ann. Internat. Mtg: SEG, 183-186.
Gregg, M. E. and Bukowski, C. T., Jr., 2000, Developing an exploration tool
in
a mature trend: A 3-D AVO case study in south Texas: The Leading Edge, 19,
no. 11, 1174-1183.
Jan 2008

342

References (cont)
Hampson, D., Russell, B. and Cardamone, M., 2004, Uncertainty in AVO-How
can we measure it?: Recorder, 29, no. 3, 5-11.
Hampson, D., and Russell, B., 1990, AVO inversion: theory and practice:
60th Ann. Internat. Mtg., Soc. Expl. Geophys., Expanded Abstracts, 14561458.
Jones, I., Christiansen, R. , Haynes, J., Faragher, J., Novianti, I., Morris, H.
and Pickering, G., 2005, Multi-Discplinary Geoscience: The 'Brenda' North
Sea Development: Recorder, 30, no. 1, 38-43.
Keho, T., Lemanski, S., Ripple, R. and Tambunan, B. R., 2001, The AVO
hodogram: Using polarization to identify anomalies: The Leading Edge,
20, no. 11, 1214-1219.
Mahob, P.N. and Castagna, J. P., 2003, AVO polarization and hodograms:
AVO strength and polarization product: Geophysics, 68, 849-862.
Mavko, G., Mukerji, T., and Dvorkin, J., 1998, The rock physics handbook
Tools for seismic analysis in porous media, Cambridge University Press.
Mavko, G., Chan, C. and Mukerji, T., 1995, Fluid substitution: Estimating
changes in Vp without knowing Vs: Geophysics, 60, 1750-1755.
O'Brien, J., 2004, Interpreter's CornerSeismic amplitudes from low gas
saturation sands: The Leading Edge, 23, no. 12, 1236-1243.

Jan 2008

343

References (cont)
Ostrander, W.J., 1984, Plane-wave reflection coefficients for gas sands at
non-normal angles of incidence: Geophysics, 49, 1637-1648.
Ratcliffe, A. and Roberts, G., 2003, Robust, automatic, continuous velocity
analysis, 73rd Ann. Internat. Mtg.: Soc. of Expl. Geophys., 2080-2083.
Richards, P.G., and Frasier, C.W., 1976, Scattering of elastic waves from
depth-dependent inhomogeneities: Geophysics, 41, 441-458.
Ross, C. P., 2002, Comparison of popular AVO attributes, AVO inversion,
and calibrated AVO predictions: The Leading Edge, 21, no. 3, 244-252.
Ross, C. P., 2000, Effective AVO crossplot modeling: A tutorial: Geophysics,
65, 700-711.
Ross, C. P. and Kinman, D. L., 1995, Non-bright-spot AVO: Two examples:
Geophysics, 60, 1398-1408.
Ross, C. P., 1995, Improved mature field development with 3D/AVO
technology:
First Break, 13, no. 04, 139-145.
Ruger, A., 1997, P-wave reflection coefficients for transversely isotropic models
with vertical and horizontal axis of symmetry: Geophysics, 62, 713-722.
Russell, B., Ross, C. P. and Lines, L., 2002, Neural networks and AVO: The
Leading Edge, 21, no. 3, 268-277.
Jan 2008

344

References (cont)
Rutherford, S.R., and Williams, R.H., 1989, Amplitude-versus-offset
variations
in gas sands: Geophysics, 54, 680-688.
Shuey, R.T., 1985, A simplification of the Zoeppritz equations: Geophysics, 50,
609-614.
Simmons, J. L., Jr. and Backus, M. M., 1994, AVO modeling and the locally
converted shear wave: Geophysics, 59, 1237-1248.
Smith, G. C. and Gidlow, P. M., 1987, Weighted stacking for rock property
estimation and detection of gas: Geophys. Prosp., EAGE., 35, 993-1014.
Thomsen, L., 1986, Weak elastic anisotropy : Geophysics, 51, 1954-1966.
Verm, R. and Hilterman, F., 1995, Lithology color-coded seismic sections:
The calibration of AVO crossplotting to rock properties: The Leading Edge, 14,
no. 08, 847-853.
Wiggins, R., Kenny, G.S., and McClure, C.D., 1983, A method for
determining and displaying the shear-velocity reflectivities of a
geologic
formation: European patent Application 0113944.
Yu, G., 1985(b), Offset-amplitude variation and controlled amplitude
processing: Geophysics, 50, 2697-2708.
Zoeppritz, K., 1919, Erdbebenwellen VIIIB, On the reflection and
propagation of seismic waves: Gottinger Nachrichten, I, 66-84.

Jan 2008

345

También podría gustarte