You are on page 1of 5

GEOPHYSICS, VOL. 67, NO. 1 (JANUARY-FEBRUARY 2002); P. 6367, 6 FIGS.

10.1190/1.1451337

Extended elastic impedance for fluid and lithology prediction

David N. Whitcombe , Patrick A. Connolly , Roger L. Reagan ,


and Terry C. Redshaw
, shear modulus , and density could be expressed in terms
of the AVO parameters A, B, and C as defined by Aki and
Richards (1980). Goodway et al. (1997) highlighted the use of
shear modulus and Lames
constant, (or, more strictly,
and ) as optimum lithology and fluid indicators.
These previous works were all formulated in the reflectivity
domain. Connolly (1999) introduced elastic impedance (EI) as
a generalization of acoustic impedance (AI) for nonnormal incidence angle, enabling the benefits of inversion to be exploited
for AVO data. He described derivations from both the two- and
three-term Zoeppritz linearizations (Aki and Richards, 1980)
using the commonly used A, B, and C parameters:

ABSTRACT

Constant angle projections of seismic sections can be


designed to provide maximum discrimination between
fluids or lithologies. The optimum projection for a noisefree, isotropic environment can be obtained using an extension to the elastic impedance concept, which itself
is an extension of acoustic impedance (AI) to nonzero
angles of incidence. To achieve this, we modify the definition of elastic impedance (EI) beyond the range of
physically meaningful angles by substituting tan for
sin2 in the two-term reflectivity equation. The primary
variable now becomes rather than . We allow it to
vary between 90 and +90 , which gives an extension
of EI for any combination of intercept and gradient. We
refer to this form of elastic impedance as extended elastic
impedance (EEI).
In this paper we demonstrate that EEI can be tuned using different values to be approximately proportional
to a number of elastic parameters, and we give EEI expressions for shear impedance (SI), bulk modulus, shear
modulus, Lames
parameter, and V p /Vs . This leads to the
identification of different areas of EEI space that tend
to be optimum for fluid and lithology imaging. Having
identified an appropriate value, the equivalent seismic
section can be obtained from combinations of intercept
and gradient stacks from routine AVO processing.

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

(1)

where is the average of the incidence and transmission angles


at a plane reflecting interface.
Connolly also showed examples of its use to provide a framework for calibrating and inverting angle stacks. He showed that
the two-term formulation of EI could be expressed as a simple
function of V p ,Vs , and density (, , and ):

EI( ) = a b c ,

(2)

where

a = (1 + sin2 ),
b = 8K sin2 ,

(3)

c = (1 4K sin ),
2

and where K is a constant, usually set to the average value of


(/)2 over the log interval of interest. This technology was
developed and applied as a fluid imaging tool in the appraisal
and development of the Foinaven field, West of Shetlands.
The EI function [equation (2)] was recently modified
(Whitcombe, 2002) by introducing reference constants o ,
o , and o , which remove the variable dimensionality of

INTRODUCTION

The present work builds on several key pieces of work. Smith


and Gidlow (1987) first showed that prestack data could be
stacked with different weights to produce their fluid factor and
pseudo-Poissons ratio sections for predicting fluids and lithology. Later, Dong (1996) showed how changes in bulk modulus

Presented at the 70th Annual Meeting, Society of Exploration Geophysicists. Manuscript received by the Editor December 5, 2000; revised
manuscript received May 30, 2001.

BP, GFU Business Unit, Burnside Road, Farburn Industrial Estate, Dyce, Aberdeen AB21 7PB, U.K. E-mail: whitcodn@bp.com.
BP, Angola Business Unit, Compass Point, 79-87 Kingston Road, Knowle Green, Staines, Middlesex TW18 1DY, U.K. E-mail: connolpa@bp.com.

BP, 501 Westlake Park Boulevard, Houston, Texas, 77079-2696. E-mail: reaganrl@bp.com.
BP Research Centre, Chertsey Road, Sunbury Upon Thames, Middlesex TW16 7LN, U.K. E-mail: redshawtc@bp.com.

c 2002 Society of Exploration Geophysicists. All rights reserved.


63

Downloaded 07 Apr 2010 to 210.212.83.82. Redistribution subject to SEG license or copyright; see Terms of Use at http://segdl.org/

64

Whitcombe et al.

equation (2) and provide an EI function which returns normalized impedance values for all angles :

"

EI() = o o

c #

(4)

The present work integrates the previous reflectivity domain


formulations with EI concepts and enables EI to be extended
to identify both fluids and lithology.
THEORY

Dong (1996) showed that the linearized Zoeppritz equations


lead to the following approximation:

1 =

(3A + B + 2C) 2
,
1.5

(5)

where 1 is the change in bulk modulus across the interface.


Various authors have noted the difficulty of extracting the C
term from real seismic data. Shuey (1985) examined the ratio
C/A and noted that this parameter tended to lie between 0
and 1. We define C/A as f . The value of f = 0.8 describes
rocks that follow Gardner et al.s (1974) relationship. Rather
than determine C directly, we replace it with f A and select an
f value appropriate to the rock properties in the area.
Using this substitution and dividing both sides of equation (5) by , the average bulk modulus across the interface,
allows us to determine bulk modulus reflectivity R :

R =

1
2

A+

B
3+2f

3+2f
.
3 4K

(6)

The first bracketed term can be considered in the form of a


first-order AVO equation, namely

R() = A + B sin2 .

(7)

Therefore,

sin2 =

1
.
3+2f

(8)

Setting f = 0.8 (Gardners relationship) results in sin2 =


0.22, or = 28 . The value R is therefore approximately proportional to R(28 ). Note that f = 0.0 and f = 1.0 provide
sin2 values of 0.33 and 0.20 (equivalent of 35 and 26 ,
respectively).
The second bracketed term in equation (6) can be considered a scaling factor. If K and f are constant over the interval
of interest, this term will be a constant. For typical values of
f = 0.8 and K = 0.25, this scaling factor equals 2.3.
Thus, the reflectivity associated with a bulk modulus log can
be considered as a scaled AVO projection.
A similar treatment (C. Sondergeld, 1999, personal communication) for Lames
constant yields

1 = (2A + B + C) 2 .

R =

1
2

B
A+
2+ f

2+ f
.
2 4K

sin2 =

1
.
2+ f

(11)

Using Gardners relationship gives sin2 = 0.36 or = 37 .


Thus, R is approximately proportional to R(37 ).
The second bracketed term in equation (10) can be considered a scaling factor. If K and f are constant over the interval
of interest, this term will be a constant. For typical values of
f = 0.8 and K = 0.25, this scaling factor will equal 2.8.
These observations are consistent with the observation that
fluids often better image at far offsets, and they imply that
the elastic impedance equation can deliver impedance proportional to bulk modulus and Lames
constant.
For lithological imaging we look to shear modulus . This
is attractive because, for rocks that follow Gassmanns (1951)
assumptions, shear modulus will be independent of fluid fill.
Following Dong (1996),

1 =

(C B) 2
.
2

(12)

Dividing both sides of equation (12) by , the average shear


modulus across the interface, allows us to determine shear
modulus reflectivity R :

R =

1
2

B
f

f
.
4K

(13)

Thus, as for bulk modulus and Lames


parameter, the reflectivity associated with a shear modulus log can be considered a scaled AVO projection. For typical values of f = 0.8
and K = 0.25, the scaling factor will equal 0.8.
The projection angle is defined by

sin2 =

1
.
f

(14)

For f = 0.8, a value for sin2 of 1.25 is obtained. In this case


there is no real solution for , though we can clearly combine
A and B together with any weights we choose. The two-term
AVO linearization defines a straight line for reflectivity against
sin2 (). This line can be extended infinitely in either direction,
as indicated in Figure 1. It only approximates physical reality
from zero to about 30 , as outlined by Shuey (1985), and has

(9)

Dividing both sides of this equation by , the average Lames

constant across the interface, allows us to determine Lames

constant reflectivity R :

Again, the first bracketed term can be considered in the form of


a first-order AVO equation with a projection angle defined
by

(10)

FIG. 1. Prestack amplitude observations are fit with the


two-term AVO linearization. The linear model can be extrapolated beyond the range of measured data and beyond the range
over which sin2 is physically definable.

Downloaded 07 Apr 2010 to 210.212.83.82. Redistribution subject to SEG license or copyright; see Terms of Use at http://segdl.org/

Extended Elastic Impedance

an equivalent angle for values of sin2 between 0 and 1. For


values of sin2 less than zero or greater than unity, there is no
equivalent real angle. But we can measure A and B from real
data between zero and 30 and hence construct reflectivity by
extrapolation for any value of sin2 that we choose.
EXTENDED ELASTIC IMPEDANCE

65

"
EEI() = AIo

AI
AIo

cos( )

GI
AIo

sin( )

#
.

(22)

To reiterate, our intent is not to produce a model that replicates observed reflectivity beyond 30 and up to the critical
angle. We are defining a useful model that can be constructed

Our objective is to express the preceding reflectivity equations in terms of the corresponding impedance relationships.
There are two difficulties with using the current EI definition.
There is the requirement for |sin2 | to exceed unity, and reflectivity values may exceed unity as sin2 increases; clearly, no
impedance contrast can give rise to a reflectivity value greater
than unity (unless we allow negative impedance). In practice,
this will mean that as |sin2 | approaches and passes unity,
the EI log, by its current definition, will become increasingly
inaccurate.
To compensate for these difficulties, we make two changes to
the current definition of EI. First, we replace sin2 by tan so
the equation is defined between rather than the 01 limit
imposed by sin2 . We also define a scaled version of reflectivity
to be normal reflectivity multiplied by cos , which ensures that
reflectivity never exceeds unity.
The first substitution in the two-term linearized Zoeppritz
equation gives

R = A + B tan ,

(15)

from which we derive

R=

(A cos + B sin )
.
cos

(16)

FIG. 2. The EEI functions for various values for well


204/24a-2. Note the inverse correlation between EEI( = 90 )
and EEI( = 90 ).

Now introducing R S , or scaled reflectivity,

R S = R cos ,

(17)

R S = A cos + B sin .

(18)

results in

The EI equivalent of equation (18) is then

"

EEI() = o o

r #
,

(19)

where

p = (cos + sin ),
q = 8K sin ,

(20)

r = (cos 4K sin ).
We call this extended elastic impedance, or EEI.
Scaled reflectivity has the property that it ranges from a value
of A at = 0 to a value of B at = 90 . The EEI equivalent to = 0 is of course acoustic impedance and at = 90
EEI will have a reflectivity corresponding to B. We refer to
EEI( = 90 ) as gradient impedance or GI.
Alternatively, by defining

AIo = o o ,
equation (19) can be written as

(21)

FIG. 3. Comparisons between elastic parameters and equivalent EEI curves for well 204/24a-2, showing the high degree of
correlation. The EEI function is defined as a function of the
angle , not the reflection angle .

Downloaded 07 Apr 2010 to 210.212.83.82. Redistribution subject to SEG license or copyright; see Terms of Use at http://segdl.org/

66

Whitcombe et al.

for any real linear combination of A and B, effectively extrapolating the observations along the sin2 axis in either direction
beyond the physically observed range. We will show that this
approach allows us to define a single function that is propor-

tional to several different elastic parameters, depending on the


value of used.
Figure 2 shows an example from the West of Shetlands well
204/24a-2, where we display EEI across the spectrum of
values (90 to +90 ). Note that the = 90 log is proportional to the inverse of the = +90 log, as expected from
equation (22).
We can therefore now give expressions for the elastic constants directly in terms of EEI by substituting the values for
sin2 in terms of f as given in equations (8), (11), and (14). By
assuming values for f of 0.8 (Gardner) and 0.25 for K , these
equations can be simplified as follows:

EEI( = 12.4 ) EI( = 28 ),


EEI( = 19.8 ) EI( = 37 ),
EEI( = 51.3 ) (no equivalent real value of ).
In addition to the rock parameters already considered, equation (19) also predicts, for K = 0.25, relationships between EEI
and shear impedance (= ) and the ratio /:
and

EEI( = 45 ) ()1.414

(23)

1.414

EEI( = 45 )
.

(24)

FIG. 4. The correlation coefficients between EEI and an Sw and


a gamma-ray curve for a range of values of .

So, including these with K , , , and AI, we have defined a


function that can be made to be approximately proportional
to six elastic parameters by adjusting a single variable . The
equation allows for expressions that are interpolations between
any of these parameters, which lets us fine-tune any property
we require.
EXAMPLES

FIG. 5. Comparison of the optimized EEI curves with the target


Sw and gamma-ray logs.

Several approximations have been introduced to reach this


point: the replacement of C by a scaled version of A, fixing K as
constant, and the approximations inherent in the Zoeppritz linearizations. To test their validity, an example from the 204/24a-2
well is used. In these examples f = 0.76 and K = 0.21 were
used. In each case the correlation coefficient between the elastic parameter and the EEI curve was 0.97. Figure 3 displays
(/), , K shear impedance, and against their equivalent

FIG. 6. Maps generated using EEI data sets using values tuned to optimize the imaging of lithology ( = 51.3 ) and fluids
( = 12.4 ), respectively.

Downloaded 07 Apr 2010 to 210.212.83.82. Redistribution subject to SEG license or copyright; see Terms of Use at http://segdl.org/

Extended Elastic Impedance

calculated EEI logs. Note the high degree of correlation in


all cases. The high correlation between (/) and EEI is perhaps surprising, given the constant K assumption [i.e., constant
(/)2 ] in the EI and EEI formulation.
FLUID AND LITHOLOGY IMPEDANCE

Bulk modulus and Lames


parameter tend to lie within an
area of EEI space with values of from about 10 to 30 ,
and shear modulus lies within a range of from about 30 to
90 . These areas are therefore likely to be good starting points
to look for optimum fluid and lithology impedance functions,
respectively. One way to search for the optimum EEI function
is to maximize the correlation between the EEI curve and a
target curve such as water saturation Sw or gamma for a range
of values.
Figure 4 show the results of correlating a range of EEI
curves with an Sw and gamma-ray curve. All curves have lowfrequency trends removed. For these data the highest correlation with the Sw curve is at = 35 , higher than the bulk
modulus or Lames
parameter values. The lithology correlation
coefficient peaks at a positive value for = 71 , with large negative values of also having high correlation. Figure 5 shows
the Sw and gamma-ray curves, together with the EEI curves
which optimally correlate. In practice, noise in the seismic section, the presence of anisotropy, velocity errors, etc., may mean
that the optimum angle to use with real data may be different
from that shown by using these log plots, but the degree of
correlation indicates what is possible with good-quality data.
As a second example, in Figure 6 we show maps of average
EEI averaged over a 25-ms time gate measured from top reservoir for the Forties field in the central North Sea. The images are
tuned to shear and bulk modulus. These have been obtained by
using f = 0.8, giving angles of 51.3 and 12.4 (sin2 values
of 1.25 and 0.22, respectively). Here we are clearly imaging
the channel systems and, potentially, remaining hydrocarbons
within these channels.

67
CONCLUSIONS

We have introduced an extension of the EI concept and


have demonstrated that, by adjusting a single parameter ,
EEI can provide a good approximation to acoustic impedance,
bulk modulus, Lames
parameter, / ratio, shear impedance,
and shear modulus and can be optimized as a fluid or lithology discriminator. Having established a desired value of , an
equivalent seismic section can be constructed using conventional impedance inversion and AVO processing techniques.
ACKNOWLEDGMENTS

We thank Keith Nunn for his comments, which improved


the submitted manuscript. We also thank the board of BP for
permission to publish this paper.
REFERENCES
Aki, K. I., and Richards, P. G., 1980, Quantitative seismology: W. H.
Freeman & Co.
Connolly, P., 1999, Elastic impedance: The Leading Edge, 18, No. 4,
438452.
Dong, W., 1996, A sensitive combination of AVO slope and intercept
for hydrocarbon indication: 58th Conference and Technical Exhibition, Eur. Assn. Geosci. Eng., paper M044.
Gardner, G. H. F., Gardner, L. W., and Gregory, A. R., 1974, Formation
velocity and densityThe diagnostic basics for stratigraphic traps:
Geophysics, 39, 770780.
Gassmann, F., 1951, Uber die Elastizitat Poroser Medien: Vierteljahrschrift der Naturforschenden Gesellschaft in Zurich, 96, 123.
Goodway, B., Chen, T., and Downton, J., 1997, Improved AVO fluid
detection and lithology discrimination using Lame petrophysical parameters, , , and / fluid stack, from P and S inversions: National convention, Can. Soc. Expl. Geophys., Expanded
Abstracts, 183186.
Shuey, R. T., 1985, A simplification of the Zoeppritz equations: Geophysics, 50, 609614.
Smith, G. C., and Gidlow, P. M., 1987, Weighted stacking for rock
property estimation and detection of gas: Geophys. Prosp., 39, 915
942.
Whitcombe, D. N., 2002, Elastic impedance normalization: Geophysics,
67, 6062, this issue.

Downloaded 07 Apr 2010 to 210.212.83.82. Redistribution subject to SEG license or copyright; see Terms of Use at http://segdl.org/

You might also like