-
The inverse scattering series for tasks associated with
primaries: Depth imaging anddirect non-linear inversion of 1D
variable velocity and density acoustic mediaHaiyan Zhangand Arthur
B. Weglein, University of Houston
Summary
This paper presents the rst analysis, and direct depthimaging
and inversion algorithms for tasks associatedwith primaries for a
vertically varying velocity anddensity acoustic medium. The method
is derived fromthe inverse scattering series, and, hence, assumes
theactual subsurface properties governing the propagationof waves
is neither known nor determined. The terms inthe series that
correspond to tasks for imaging-only andinversion-only are identied
and separated. Tests withanalytic data indicate signicant added
value, beyondlinear estimates, in terms of both the proximity to
actualproperties and the increased range of angles over whichthe
improved estimates are valid.
Introduction
The ultimate objective of inverse problems is to deter-mine
medium and target properties from reectiondata. However, that
objective has never been achievedin a straight ahead single step
manner. The onlydirect multi-dimensional inversion procedure for
seismicapplication, is the task specic subseries of the
inversescattering series. (Weglein et al., 2003). The order
ofprocessing tasks is (1) free surface multiple removal;(2)
internal multiple removal; (3) imaging primariesin space and (4)
inversion. The free surface multipleremoval and internal multiple
attenuation subseries havebeen presented by Weglein et al. (1997).
Those twomultiple removal procedures are model type
independent.Taking internal multiple algorithm from attenuationto
elimination is being studied (Ramirez and Weglein,2005). In
contrast with model type independent multipleremoval procedures,
there is a full expectation that tasksand algorithms associated
with primaries will have acloser interest in model type. For
example, there is noway to even imagine that medium property
identicationcan take place without reference to a specic model
type.A staged approach and isolation of tasks philosophy
isessential in this yet tougher neighborhood of primaries.The
stages within the strategy for primaries are asfollows: (1) 1D
earth, with one parameter, velocity as afunction of depth, and a
normal incidence wave (Shaw etal., 2003); (2) 1D earth with one
parameter subsurfaceand oset data, one shot record (Shaw, 2005);
(3) 2Dearth with one parameter, velocity, varying in x and z,and a
suite of shot records (Liu et al., 2005); (4) 1Dacoustic earth with
two parameters varying, velocityand density, one propagation
velocity, and one shotrecord of PP data, and (5) 1D elastic earth,
two elastic
1617 Science & Research Bldg 1, Houston, TX 77204-5005
isotropic parameters and density, and two wave speeds,for P and
S waves, and PP, PS, SP, and SS shot recordscollected (Zhang et
al., 2005). This paper deals withstage (4). The model is acoustic
and a second paper inthis progression and evolution generalizes
this for theelastic case.
In this paper, the rst direct non-linear inversion methodis
obtained for 1D two-parameter acoustic media and a2D experiment.
From this solution, the tasks for imaging-only and inversion-only
terms are successfully separated.Tests with analytic data indicate
signicant added value,beyond linear estimates, in terms of both the
proximity toactual value and the increased range of angles over
whichthe improved estimates are provided. The depth
imagingalgorithm without the velocity rst decides if the
inputvelocity is adequate, and if adequate conventional migra-tion
is prescribed. If the data decides the verdict on thevelocity is
inadequate it acts to remove the incorrect im-age and constructs
the correct one, without knowing ordetermining the velocity. Other
terms in the algorithmare identied as performing non-linear direct
AVO, andonce again, allow the data self determination of
overbur-den velocity adequacy, and acts accordingly to improveupon
linear estimates of property changes. The role of ve-locity is
claried, as central and all-important in location,in that only an
incorrect velocity causes a depth imagingresponse from the series,
independent of how you param-eterize the acoustic problem, or what
error estimates ofother properties might suer. Benet of non-linear
directinversion is demonstrated over linear standard procedure,for
a set of examples using analytic and numerical tech-niques. The
common problem of linear leaking betweenlinear property change
predictions is also addressed by theseries, and located and
analyzed in this paper.
The paper has the following structure: In section 2 weshow the
derivation in detail and then discuss a specialparameter followed
by numerical tests. The last sectionhas conclusions and
acknowledgements.
Derivation of 1, 1 and 2, 2
To illustrate task (4), target identication, we considera 1D
acoustic two-parameter earth model (e.g. bulkmodulus and density or
velocity and density). We startwith the 3D acoustic wave equations
in the actual andreference medium (Clayton and Stolt, 1981 and
Wegleinet al., 1997)
[2
K(r)+ 1
(r)]G(r, rs;) = (r rs), (1)
SI 3.4
SEG/Houston 2005 Annual Meeting 1705
Dow
nloa
ded
03/2
0/15
to 1
29.7
.0.9
4. R
edist
ribut
ion
subje
ct to
SEG
licen
se or
copy
right;
see T
erms o
f Use
at htt
p://lib
rary.s
eg.or
g/
-
Direct non-linear two parameter 2D acoustic inversion[2
K0(r)+ 1
0(r)]G0(r, rs;) = (r rs), (2)
where G(r, rs;) and G0(r, rs;) are respectively thecausal Greens
operators that describe wave propagationin the actual and reference
medium. The bulk modulus,K, is given by c2, where c is P-wave
velocity and is thedensity. Then the perturbation can be written as
(We-glein et al., 2003)
V =2
K0+
0, (3)
where = 1 K0K
and = 1 0
are the two parameters
we choose to perform the following non-linear inversion.Assuming
both 0 and c0 are constants, and for a 1-Dacoustic medium, we
expand V (z,), (z) and (z) inthe same way as (Weglein et al.,
1997)
V = V1 + V2 + , (4)
we have
V1(z,) = 21(z)
K0+
1
01(z)
2
x2+
1
0
z1(z)
z, (5)
V2(z,) = 22(z)
K0+
1
02(z)
2
x2+
1
0
z2(z)
z, (6)
...,
where the subscript i in Vi, i and i denote the portionof those
quantities i-th order in the data. Substituting(5) into the rst
equation of the inverse scattering series(Weglein et al., 2003) D =
[G0V1G0]ms, we can obtainthe linear solution for 1 and 1 in the
frequency domain
D(qg, , zg, zs) = 04
eiqg(zs+zg)
[
1
cos2 1(2qg) + (1 tan2 )1(2qg)
], (7)
where the subscripts s and g denote source and
receiverquantities respectively, and qg, and k = /c0 shownin Figure
1, have the following relations (Matson, 1997):qg = qs = k cos , kg
= ks = k sin . Similarly, substitut-
gq k
gk
z
000,, Kc
Kc ,,
Fig. 1: The relationship between qg , kg and .
ing (6) into the second equation of the inverse scattering
series [G0V2G0]ms = [G0V1G0V1G0]ms (Weglein et al.,2003), we
derive the solution for 2(z) and 2(z) as afunction of 1(z) and
1(z)
1
cos2 2(z) + (1 tan2 )2(z)
= 12 cos4
21(z) 12(1 + tan4 )21(z) +
tan2
cos2 1(z)1(z)
12 cos4
1(z)
z0
dz[1(z) 1(z)]
+1
2(tan4 1)1(z)
z0
dz[1(z) 1(z)], (8)
where 1(z) =d1(z)
dz, 1(z) =
d1(z)dz
.
We have obtained in equation (8) the rst two parameterdirect
non-linear inversion of 1D acoustic media for a 2Dexperiment.
Equations (7) and (8) imply that two dier-ent angles , can
determine 1, 1 and then 2, 2. Fora single-interface example, we can
also show that onlythe rst three terms on the right hand side
contributeto amplitude correction, while the last two terms
per-form imaging in depth. Therefore, in this way, the tasksfor
imaging-only and inversion-only terms are identiedand examined. If
another choice of free parameter otherthan (e.g., or kh) was
selected, then the functionalform between the data and the rst
order perturbationchanges. Furthermore, the relationship between
the rstand second order perturbation is then also dierent, andnew
analysis would be required for the purpose of iden-tifying specic
task separated terms. In our experience,the choice of as free
parameter (for a 1D medium) isparticularly well suited for allowing
a task separated iden-tication of terms in the inverse series.
Details about thesignicance of this solution will be presented in
the fol-lowing sections.
Special parameter for linear inversion
As we mentioned above, since the relationship betweendata and
target property changes is non-linear, linearinversion will produce
errors in target property pre-diction. When an actual property
change is zero, thelinear prediction of the change can be non-zero.
Also,when the actual change is positive, the predicted
linearapproximation can be negative. There is a specialparameter
for linear inversion of acoustic media, thatnever suers the latter
problem the P-wave velocity.
From the one interface case, we can show that when c0 =c1, i.e.,
when c = 0, (c)1 = 0. This generalizes towhen c > 0, then (c)1
> 0, or when c < 0, then(c)1 < 0, as well.
Therefore, we can rst obtain the right sign of relativechange in
P-wave velocity from the linear inversion (c)1,then, we can obtain
a more accurate amplitude when weinclude non-linear terms. The
linear approximation tothe change in velocity,
(cc
)1, is 1
2(1 1).
SI 3.4
SEG/Houston 2005 Annual Meeting 1706
Dow
nloa
ded
03/2
0/15
to 1
29.7
.0.9
4. R
edist
ribut
ion
subje
ct to
SEG
licen
se or
copy
right;
see T
erms o
f Use
at htt
p://lib
rary.s
eg.or
g/
-
Direct non-linear two parameter 2D acoustic inversion
We also note that when the velocity doesnt change acrossan
interface, i.e., c0 = c1, looking at the integrand ofimaging terms
1 1 (see (8)), the image doesnt movebecause 1 1 = 0 in this
situation. We can see thismore explicitly when we change the two
parameters and to c
cand , and we can rewrite (8) as
1
cos2
(c
c
)2
(z) + 2(z)
=cos2 22 cos4
(c
c
)21
(z) 1221(z)
1cos4
(c
c
)1
(z)
z0
dz(
c
c
)1
1cos2
1(z)
z0
dz(
c
c
)1
. (9)
This equation makes two key points. One is that there isno
leakage correction at all in this expression. (Leakage isa two
parameter eect and its removal is associated witha cross term, like
the 11 term in equation (8)). Theabsence of a cross term in
equation (9) is an indicationof the special leakage resistant
properties of the P-wavevelocity.
The second point is that, when we look at the integrand(cc
)1of the imaging terms, it indicates that if we have
the right velocity, the imaging terms will automaticallybe zero
and no integration is performed. On the otherhand, if we didnt have
the right velocity, these imagingterms would move the interface
from the wrong locationcloser to the right location. The conclusion
is that thedepth imaging terms depend only on the velocity
errors.
Numerical tests
Consider the one interface example (in Figure 2), and
000,, Kc
111,, Kc
z
x
a
0
Fig. 2: 1D one interface acoustic model.
assume the interface surface is at z = a, and supposezs = zg =
0. The reection coecient has the followingform (Keys, 1989)
R() =(1/0)(c1/c0)
1 sin2
1 (c21/c20) sin2
(1/0)(c1/c0)
1 sin2 +
1 (c21/c20) sin2 ,
(10)where , the angle of P-wave incidence, is same as thatin
equation (7) and (8), and 0 and 1 denote the density
for the reference medium and actual medium respectively.Then
using perfect data (Clayton and Stolt, 1981 andWeglein et al.,
1986)
D(qg, ) = 0R()e2iqga
4iqg, (11)
and substituting (11) into (7), after Fourier transforma-tion
over 2qg, for z > a and xed , we obtain
1
cos2 1(z) + (1 tan2 )1(z) = 4R()H(z a). (12)
In this case of a single reector, the non-linear equation(8)
reduce to
1
cos2 2(z) + (1 tan2 )2(z)
= 12 cos4
21(z) 12(1 + tan4 )21(z)
+tan2
cos2 1(z)1(z), (13)
From equations (12) and (13) we choose two dierent an-gles to
solve for 1 and 1, and then 2 and 2.
For a specic model, 0 = 1.0g/cm3, 1 = 1.1g/cm
3, c0 =1500m/s and c1 = 1700m/s. In the following gures,
wepresent the results for the relative changes in the P-wave
bulk modulus( = K
K
), density
( =
), impedance(
II
)and velocity
(cc
)corresponding to dierent pairs
of 1 and 2.
From Figure 3, we notice that when we add 2 to 1, theresult is
much closer to the exact value of . Furthermore,the result is
better behaved, i.e., the plot surface becomesatter, over a larger
range of precritical angles. Similarly,from Figure 4, we can also
see the results of 1 + 2 aremuch better than those of 1. In
addition, we observe thatthe sign of the linear approximation to
the relative changein density, 1, is wrong at some angles, while,
the resultsfor 1 + 2 always have the right sign. After including2,
we correct the sign of the density change, which isvery important
in the generalization to identication ofan elastic earth.
Conclusion
Including terms beyond linear in the earth property
iden-tication subseries provides added value. Although themodel we
used in the numerical test is simple, equations(7) and (8) are also
generalizable for multidimensionalmedia and complex targets. The
inverse scatteringseries is a direct inversion procedure which
inverts dataindependent of the properties of the target,
withoutassumptions such as smooth geometry or small contrast.This
work is a major step towards the realism for targetidentication.
The numerical results are encouragingand this work will be extended
to study the elastic caseusing three parameters (see, e.g., Boyse,
1986 and Boyseand Keller, 1986).
SI 3.4
SEG/Houston 2005 Annual Meeting 1707
Dow
nloa
ded
03/2
0/15
to 1
29.7
.0.9
4. R
edist
ribut
ion
subje
ct to
SEG
licen
se or
copy
right;
see T
erms o
f Use
at htt
p://lib
rary.s
eg.or
g/
-
Direct non-linear two parameter 2D acoustic inversion
Acknowledgements
We would like to thank R. Keys, D. Foster, R. Stolt, B.Nita, S.
Shaw and K. Innanen for valuable discussionsand useful suggestions.
The support of the sponsorsof M-OSRP and the support from NSF-CMG
awardnumber DMS-0327778 are gratefully acknowledged.
References
Boyse W E 1986 Wave propagation and inversion inslightly
inhomogeneous media p 40
Boyse W E and Keller J B 1986 Inverse elastic scatteringin three
dimensions J. Acoust. Soc. Am. 79 215218
Clayton R W and Stolt R H 1981 A Born-WKBJ inversionmethod for
acoustic reection data for attenuatingmultiples in seismic reection
data Geophysics 4615591567
Innanen K A 2003 Methods for the Treatment of Acous-tic and
Absorptive/Dispersive Wave Field Measure-ments Ph.D. thesis
University of British Columbia
Keys R G 1989 Polarity reversals in reections from lay-ered
media Geophysics 54 900905
Liu F, Weglein A B, Innanen K A and Nita B G 2005Inverse
scattering series for vertically and laterallyvarying modia:
application to velocity independentdepth imaging M-OSRP Annual
Report 2004 176-263preparing for publication
Matson K H 1997 An inverse scattering series methodfor
attenuating elastic multiples from multicompo-nent land and ocean
bottom seismic data Ph.D. the-sis University of British Columbia p
18
Ramirez A C and Weglein A B 2005 An inverse scatteringinternal
multiple elimination method: Beyond atten-uation, a new algorithm
and initial tests M-OSRPAnnual Report 2004 138-157 preparing for
publica-tion
Shaw S A, Weglein A B, Foster D J, Matson K H and KeysR G 2003
Isolation of a leading order depth imag-ing series and analysis of
its convergence properties,M-OSRP Annual Report 2002 157-195
preparing forpublication
Shaw S A 2005 An inverse scattering series algorithmfor depth
imaging of reection data from a layeredacoustic medium with an
unknown velocity modelPh.D. thesis University of Houston
Weglein A B, Violette P B and Keho T H 1986 Usingmultiparameter
Born theory to obtain certain exactmultiparameter inversion goals
Geophysics 51 10691074
Weglein A B, Gasparotto F A, Carvalho P M and Stolt RH 1997 An
inverse-scattering series method for atten-uating multiples in
seismic reection data Geophysics62 19751989
Weglein A B, Araujo F V, Carvalho P M, Stolt R H,Matson K H,
Coates R, Corrigan D, Foster D J, ShawS A and Zhang H 2003 Inverse
scattering series andseismic exploration Inverse Problem 19
R27R83
Zhang H, Weglein A B and Keys R G 2005 Velocity in-dependent
depth imaging and non-linear direct tar-get identication for 1D
elastic media: testing andevaluation for application to non-linear
AVO, usingonly PP data M-OSRP Annual Report 2004 312-338preparing
for publication
010
2030
4050
0.16
0.18
0.20
0.22
0.24
0.26
0.28
0.30
0.32
0.34
0.36
0.38
0.40
0.42
0
10
20
30
40
50
1
theta2
theta1
010
2030
40
50
0.16
0.18
0.20
0.22
0.24
0.26
0.28
0.30
0.32
0.34
0.36
0.38
0.40
0.42
0
10
20
30
40
50
1+
2
theta2
theta1
Fig. 3: 1 (left) and 1 + 2 (right) displayed as a functionof two
dierent angles. The red line in the gures present theexact value of
. In this example, the exact value of is 0.292.
010
2030
40
50
-0.02-0.010.000.010.020.030.040.050.060.070.080.090.100.110.120.130.140.15
0.16
0
10
20
30
40
50
1+
2
theta2
theta1
010
2030
40
50
-0.02-0.010.000.010.020.030.040.050.060.070.080.090.100.110.120.130.14
0.15
0.16
0
10
20
30
40
50
1
theta2
theta1
Fig. 4: 1 (left) and 1+2 (right). In this example, the
exactvalue of is 0.09.
SI 3.4
SEG/Houston 2005 Annual Meeting 1708
Dow
nloa
ded
03/2
0/15
to 1
29.7
.0.9
4. R
edist
ribut
ion
subje
ct to
SEG
licen
se or
copy
right;
see T
erms o
f Use
at htt
p://lib
rary.s
eg.or
g/
-
EDITED REFERENCES Note: This reference list is a copy-edited
version of the reference list submitted by the author. Reference
lists for the 2005 SEG Technical Program Expanded Abstracts have
been copy edited so that references provided with the online
metadata for each paper will achieve a high degree of linking to
cited sources that appear on the Web. The inverse scattering series
for tasks associated with primaries: Depth imaging and direct
non-linear inversion of 1D variable velocity and density acoustic
media References Boyse, W. E., 1986, Wave propagation and inversion
in slightly inhomogeneous media:
40. Boyse, W. E., and J. B. Keller, 1986, Inverse elastic
scattering in three dimensions:
Journal of the Acoustical Society of America, 79 215218.
Clayton, R. W. and R. H. Stolt, 1981, A born-WKBJ inversion method
for acoustic
reflection data for attenuating multiples in seismic reflection
data: Geophysics, 46 15591567.
Innanen, K. A., 2003, Methods for the treatment of acoustic and
absorptive/dispersive wave field measurements: Ph.D. thesis,
University of British Columbia.
Keys, R. G., 1989, Polarity reversals in reflections from
layered media: Geophysics, 54 900905.
Liu, F., A. B. Weglein, K. A. Innanen, and B. G. Nita, 2005,
Inverse scattering series for vertically and laterally varying
media: Application to velocity independent depth imaging:
Mission-Oriented Seismic Research Program Annual Report,
176263.
Matson, K. H., 1997, An inverse scattering series method for
attenuating elastic multiples from multicomponent land and ocean
bottom seismic data: Ph.D. thesis, University of British
Columbia.
Ramirez, A. C., and A. B. Weglein, 2005, An inverse scattering
internal multiple elimination method: Beyond attenuation, a new
algorithm and initial tests: Mission-Oriented Seismic Research
Program Annual Report, 138157.
Shaw, S. A., 2005, An inverse scattering series algorithm for
depth imaging of reflection data from a layered acoustic medium
with an unknown velocity model: Ph.D. thesis, University of
Houston.
Shaw, S. A., A. B. Weglein, D. J. Foster, K. H. Matson, and R.
G. Keys, 2003, Isolation of a leading order depth imaging series
and analysis of its convergence properties: Mission-Oriented
Seismic Research Program Annual Report, 157195.
Weglein, A. B., F. A. Arajo Gasparotto, P. M. Carvalho, and R.
H. Stolt, 1997, An inverse-scattering series method for attenuating
multiples in seismic reflection data: Geophysics 62, 19751989.
Weglein, A. B., F. V. Arajo Gasparotto, P. M. Carvalho, R. H.
Stolt, K. H. Matson, R. Coates, D. Corrigan, D. J. Foster, S. A.
Shaw, and H. Zhang, 2003, Inverse scattering series and seismic
exploration: Inverse Problems, 19, R27R83.
Weglein, A. B., P. B. Violette, and T. H. Keho, 1986, Using
multiparameter Born theory to obtain certain exact multiparameter
inversion goals: Geophysics, 51 10691074
Zhang, H., A. B. Weglein, and R. G. Keys, 2005, Velocity
independent depth imaging and non-linear direct target
identification for 1D elastic media: testing and
Dow
nloa
ded
03/2
0/15
to 1
29.7
.0.9
4. R
edist
ribut
ion
subje
ct to
SEG
licen
se or
copy
right;
see T
erms o
f Use
at htt
p://lib
rary.s
eg.or
g/
-
evaluation for application to non-linear AVO, using only PP
data: Mission-Oriented Seismic Research Program Annual Report,
312338.
Dow
nloa
ded
03/2
0/15
to 1
29.7
.0.9
4. R
edist
ribut
ion
subje
ct to
SEG
licen
se or
copy
right;
see T
erms o
f Use
at htt
p://lib
rary.s
eg.or
g/