Вы находитесь на странице: 1из 4

Examplesof a NonlinearInversionMethod Basedon the T Matrix of

ScatteringTheory:Applicationto Multiple Suppression


E M. CarvaIho? PPPG/Federal Univ. of Bahia and Petrobras,Brazil;

SP1.4

Downloaded 03/20/15 to 129.7.0.94. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

A. B. Wegiein,ARC0 Oil and Gas; and R. H. Stolt, Conoco Inc.


are appropriate for water bottom multiples (Wiggins,
1988).
Multiple suppression is an important,
longstanding and only partially solved problem in
exploration seismology. A multidimensionalmethod
derived from inverse scattering theory (A. B. Weglein
and R. H. Stolt, 1991) invertsseismicreflectiondata and
removes multiples. The procedure does not rely on
periodicityor differentialmoveout,nor does it require a
modelof the multiplegeneratingreflectors. An estimate
of the sourcesignatureis required. The inversesolution
is expressed in a series. Each successive term in the
series requires a migration-inversionof progressively
more complicated data. However, in this nonlinear
method each migration-inversionis performedwith the
6ame constant velocity mig.@ion-inversionoperator.
That is, in contrast to iterative linear methods, the
operatorinvertedis the same for each term in the series.
Our experienceindicatesthat convergenceis rapid for a
wide class of examples. For very strongreflectorsand a
homogeneous starting model, four or five terms
produced excellent results. In principle, the method
suppressesall multiples. However,in its presentform, it
is computationally feasible for removing all surface
multiples.
lCiTRODUCTlON
Methodsfor removingmultiplereflectionshave a
long historyin explorationseismology. At the workshop
on the Suppressionor Exploitationof Multiples at the
1990 SEG annual meeting, several excellent papers
reviewed and compared these techniques. Hardy and
Hobbs (1991) also present an overview and propose a
multiplesuppressionstrategy.
NMO-stacking,f-k and p-z filtering are examples
of moveout based methods requiringstacking velocity
information. NMO-stackinggenerally is effective when
the moveout differences between a primary and a
multiplewith the same zero offset time for 10,000 ft of
offset is, after moveoutcorrection,greater than 40 ms.
Optimized p-z filtering methods usually can separate
primary from multiple when, in the same moveout
correctedCMP gather,they differby at least 10 ms.
Predictive deconvolutiontechniques depend on
multiplesbeing periodicreplicasof the primary at each
receiver. They dont depend on moveoutdifferencesOr
require stacking velocity information. However, their
shortcomingis that they can predictonly multipleswhich
are periodic. Space-time methodsare effective Onlyat
(or near) zero offsetand (near) verticalincidence. P- Z
methods in principle extend periodicityto all p-values,
but in practicehave their own problemsand limitations.
Modeling and subtraction methods reqUire an
accurate estimateof the sourceof multiplesand, hence,
1319

As expected, discontinuous reference velocity


linear inversion (e.g., Lui, 1984) suffers the same
restrictions/benefitsas wave equation modeling and
subtraction.
Ware and Aki (1969) and Verschuur,et al. (1988)
are among those using a reflectivity matrix model of
seismic data (in one and multidimensions,respectively)
to develop algorithms for the removal of surface
multiples.
RFAI ISTIC NQbllINFAR SFlSMlC INVFRSlDN GO&&
f!d!JtTlPr F SUPPRFSSU
Nonlinearinversionmethodsoften are viewed as
a procedure to improve the parameter estimation
capabilityof a linear method. A less ambitiousand more
realistic goal for nonlinear seismic inversion is the
suppressionof multiples.
In exploration geophysics, iterative linear
inversion(in all its differentforms) is commonlythought
of as the only nonlinear approach. In fact, there are
numerousalternate nonlinearapproaches. The T matrix
formalism of scattering theory provides a useful
framework for inverse scatteringproblemsand various
linear and nonlinearapproaches.
Amongearly works in this area are Moses (1956)
Wolf (1969), Razavy (1975) and Prosser (1980).
Weglein, Boyse and Anderson (1981) and Stolt and
Jacobs (1980a, 1980b) adopted and extended this T
matrixtechnologyfor the surfaceseismicproblem.

Define Green functionsG, Go, wave functionsP,


PO, and differentialoperatorsL and Lo in the actual and
reference media, respectively. A(w) is the source
signature.
The Lippman-Schwinger
equationexpressedin terms of
P and POis

P=<+G,JfP
where

V=Lo-L
Definethe T operatoras

Inversion and multiple suppression

TGos VG

Typically, the right hand sides of equations (6), (7) and


(8) require a volume integral over the subsurface for
each factor Go. For a homogeneous reference medium,
with a free surface, Go consists of two terms. The first
propagates directly from point (A) to point (B) in the
subsurface, whereas the second represents the
propagation from (A) to the free surface and then to (B).
If one ignores the first of these and retains the second,
then the right hand sides of (6), (7) and (8) become
surface integrals.
(see Stolt and Jacobs, 1980b
Weglein and Stolt, 1991) This amounts to keeping
surface multiples and ignoring interbed multiples.

(1)

Downloaded 03/20/15 to 129.7.0.94. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

From this definition and the Lippmann-Schwinger


equation, one finds

T=V+VGV
T=V+TGoV
A
A
Define
8 and
s as projections onto the receiver and
source planes, respectively.

EXAMPLES
We show test results for two one-dimensional
models. In each case our data is a shot record Fourier
transformed analytically over offset. Go is the Green
function for a homogeneous medium, with a free
surface, whose properties correspond to the layer
containing the source and receiver.

Then the seismic scattered (or reflection)data is

D=P-p,
D / A = ABGoVGAs
= hgGoTGohl (3)

Figure 1a illustrates the first model. Figures 1b


and lc give the inversion result for the V1 and Vl+V2 +
V3 + V4 terms, respectively. As expected, the surface
multiples are suppressed and the small interbed
multiples remain.

The goal is to determine V from measured values of D


on the surface of the earth.
Expand T and V in a power series in D/A

T=&T,
n=l

v = &v,
(4)

?I=1

(5)

Substitution of (4) and (5) into (2b) and (3) and equating
equal powers of s,

In the second example (Figure 2a) the moveout


difference between the primary from the lower reflector
and the first multiplefrom the upper reflector is 6 ms over
10,000 feet of offset. This is in the range (0 to 10 ms)
where differential moveout methods would have
difficulty. Figures 2b and 2c show VI and Vl+V2 for this
model. The multiple is removed in Vl+V2.

from &I

CONCLUSIONS

D / A = hgGoV,G,,As

The nonlinearity of a new wave theoretic


inversion method can be exploited for multiple
suppression. The multidimensional method does not
rely on moveout differences, periodicityor modeling.

where E is a parameter used to track the power in the


data D/A. After the calculationof T and V, E is set to one.

from c*

The procedure consists of a series of uncham&g


migration-inversionoperations applied to a sequence of
effective data. Initial synthetic data tests indicate rapid
convergence for a wide range of models. An estimate of
the source signature is required.

(7)
from c3

-mGFMFNTS
CNPQ and Petrobras are thanked for supporting
PMC and ABW during a sabbatical year in Brazil. ARC0
Oil and Gas Co. and Conoco, Inc. are thanked for
continuous support and encouragement. Doug Foster
and Tim Keho are thanked for helpful discussions and
comments.

and so on.

Accordingto equations (6), (7) and (a), each successive


:erm in the series for V is found by migration-inversion
withthe same operator G

1320

Inversion and multiple suppression

--..

Downloaded 03/20/15 to 129.7.0.94. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

PEFERENCFS

REFLECTOR 0

Weglein, A. B. and Stolt, R. H. (1991), I. The wave


physics of downward continuation, wavelet estimation,
and volume and surface scattering: II. Approaches to
linear and non-linear migration-inversion,Mathematical
Frontiers in Reflection Seismology edited by W. W.
Symes, SIAM/SEG.

FREE SURFACE

1 0.01 Km
source/receiver

2 Km/s

t
0.0; Km

REFLECTOR 1

I
t
0.072 Km

3 Km/s

Stolt, R. H. and Jacobs, B. (1980a), Inversion of seismic


data in a laterally heterogeneous medium, Stanford
Exploration Report No. 24, pp. 135-152, (1980b) An
approach to the inverse seismic problem, Stanford
Exploration Report No. 25, pp. 121-I 34.

REFLECTOR 2
4 Km/s
Figure la. ModelNumber1.

Weglein, A. B., Boyse, W. E. and Anderson, J. E. (1981),


Obtaining three dimensional velocity information directly
from reflection seismic data: An inverse scattering
formalism: Geophysics, V. 46, no. 8.

0.6

Lui, C. Y., (1984), Born inversion applied to reflection


seismology, Ph.D. Thesis, U. of Tulsa, Department of
Geophysics.

0.3

Moses, H. E. (1956), Calculation of the scattering


potential from reflection coefficients, Phys. Rev., V. 102,
pp. 559-567.

10102
10202
10201
201M
"ltMJc~020,

1o.y

-2

0.0

~1010*02

Razavy, M. (1975), Determination of the wave velocity in


an inhomogeneous medium from the reflection
coefficients, Journ. Acousti. Sot. Am., V. 58, pp. 956963.

A2
101'

-0.3

I
212

1020102
1020201
2010102

'0'0102
10mm

I\ 1010101

102

1020101

2010201

201

20lolO'

2020101

Devaney, A. J. and Weglein, A. B. (1989), Inverse


scattering using the Heitler equation. Inverse Problems,
December, 1989, V. 5, No. 3, pp. 49-52.

-0.6

Hardy, R. J. J. and Hobbs, R. W. (1991), A strategy for


multiple suppression, First Break, V. 9, No. 4, April,
1991.

Figure lb. dz

Verschuur, D. J., Herrmann, P., Kinneging, N. A.,


Wapenaar,
C.P.A. and Berkhout, A. J. (1988),
Elimination of surface-related multiply reflected and
converted waves, 58th Meeting, Society of Exploration
Geophysicists, Expanded Abstracts, pp. 1017-l 020.

0.6

*
MULTlPLES

0.0

0.2

0.4
0.6
DEPTH (Km)

0.8

1.0

dV1
for ModelNumber1.

Ware, J. A., and Aki, K. (1969), Continuous and discrete


inverse scattering problems in a stratified elastic
medium. 1. Plane waves at normal incidence. Journ.
Acoust. Sot. Am., V. 45, pp. 91 l-921.

0.3

4 0.0
Wiggins, J. W. (1988), Attenuation of CWTIpleXwaterbottom multiples by wave-equation based prediction
and subtraction, Geophysics, V. 53, pp. 1527-1539.

-0.6j--,--0.0

1321

Figure lc.

0.2

d (VI +V,+V3. )

dz

0.4
0.6
DEPTH (Km)

0.8

1.0

forModelNumber1.
-

Downloaded 03/20/15 to 129.7.0.94. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

Inversion and multiple suppression

REFLECTOR 0
t
2400 m

0.4

FREE SURFACE

V q 3600 m/s

REFLECTOR 1

0.2

1st INTERFACE

I
REFLECTOR 2

wIa
; 0.0

2nd INTERFACE

V ~4100 m/s

SGI
I

Figure2a. ModelNumber2. Thefirstmultiplefromthefirst


reflectorhasa moveoutpatternverycloseto themoveoutof the
primaryfromthesecondreflector.

-0.2

-0.4
0

4
DEPTH (Km)

d! for themodelin Figure2a.


Figure2c. dz

3600
. ..\....\..........\...........
2400

0.4

3 1800 3
2
F 1200 -

0.2
m Ia

600

+N
g

0.0

600

1200
1800
OFFSET(M)

2400

3600

-0.2

Figure2b. Moveoutpatternsfor modelin Figure2a.The


solidlineis theprimaryof thefirstreflector.Thedashed
line
is theprimaryof thesecond
reflectorandthefirstmultipleof
thefirstreflector.

-0.4
0

4
DEPTH (Km)

d(V +V )
for themodelin Figure2a.
Figure2d. 1 dz

1322

Вам также может понравиться