GEOPHYSICS, VOL. 74, NO. 4 共JULY-AUGUST 2009兲; P. S95–S104, 10 FIGS.

10.1190/1.3137018

Target-oriented curved-wave prestack depth migration

by controlled illumination

Xiumei Chen1, Zaitian Ma2, Guohua Nie1, and Huazhong Wang2

ABSTRACT get-oriented synthesis operators by wave theory, combined with

rotation or perturbation of the illuminating direction of a base-
Conventional prestack depth migration 共PSDM兲 based on full kernel synthesis operator. The base-kernel operator is obtained
prestack data involves many computations for wavefield extrap- by inverse extrapolation of a predefined source wavefield
olation. Areal shot-record technology offers an attractive alterna- at the given target level. Different model-constrained synthesis
tive for efficient PSDM because its synthesis process greatly de- operators are constructed by rotating or perturbing illumination
creases the amount of prestack data required for migration. Con- of the kernel operator. Then they are applied to shot records to
structing its synthesis operator is key to the image quality of the synthesize curved-wave records. The resulting curved wave-
migrated areal shot record. Curved-wave PSDM technology ex- fields are entered for depth extrapolation and imaging, combin-
pands areal shot-record migration. It constructs synthesis opera- ing image results from different curved wavefields to produce
tors by defining a complex function as a base kernel of curved the total image. The method controls the illumination of
wavefields multiplied by a factor of its illumination perturbation synthesized source wavefields in a target-oriented way directly at
related to ray parameter and then synthesizes curved wavefields the surface and can achieve high-quality images of the target
for migrations. Based on curved-wave-migration theory, we pro- zone with great efficiency. Numerical demonstrations on the
pose an efficient and accurate curved-wave, controlled-illumina- standard Marmousi model provide good imaging results of the
tion method to migrate in a target-oriented way. We construct tar- complicated structure.

INTRODUCTION plies it to shot records to synthesize an areal shot record. The areal
shot record is migrated with a conventional wave-equation depth-
Kirchhoff prestack depth migration 共PSDM兲 has extensive eco-
migration method suitable for shot gathers. The synthesis process
nomic application in industry, but it is inaccurate for very complex
significantly reduces the amount of prestack data and subsequently
media. Wave-equation-based PSDM is the most appropriate tech-
improves migration efficiency. Plane-wave migration is a simple
nology to image complex geologic structures accurately. A conven-
case of areal shot migration. Duquet and Lailly 共2006兲 develop an ef-
tional migration scheme based on field-shot data surveyed by point
sources requires many computations for wavefield extrapolation, es- ficient 3D plane-wave migration method applied to streamer data
pecially in three dimensions. Over the past few decades, some wave- with a combination of cylindrical-wave synthesis and a time-shift
equation depth-migration methods based on transfer of data domain technique.
or reconstruction of prestack data have acquired favorable migration Based on areal shot-record technology, Rietveld et al. 共1992兲 in-
efficiency. These include common-azimuth migration and migration troduce the concept of optimum illumination oriented to a geologic
of synthesized prestack data 共Biondi and Palacharla, 1996; Foster et target. Rietveld and Berkhout 共1994兲 further develop PSDM by con-
al., 2002兲. trolled illumination, aiming for high-quality, efficient imaging of the
Berkhout 共1992兲 introduces an efficient, accurate areal shot- target area. The desired illumination at a given target is constructed
record technology. He designs a complex synthesis operator and ap- by predefining sources at target level. Required synthesis operators

Manuscript received by the Editor 12 December 2007; revised manuscript received 18 December 2008; published online 19 June 2009; corrected version pub-
lished online 29 June 2009.
1
Tongji University, School of Aerospace Engineering and Applied Mechanics, Institute of Applied Mechanics, Shanghai, China. E-mail: xmchen@
tongji.edu.cn; ghnie@tongji.edu.cn.
2
Tongji University, School of Ocean and Earth Science, Shanghai, China. E-mail: mazaitian@tongji.edu.cn; herbhuak@vip.163.com.
© 2009 Society of Exploration Geophysicists. All rights reserved.

S95

Downloaded from http://pubs.geoscienceworld.org/geophysics/article-pdf/74/4/S95/3220115/gsgpy_74_4_S95.pdf

by CNOOC user
on 31 August 2022
 S96 Chen et al.

are calculated by inverse extrapolations of predefined source wave- P共z0兲 ⳱ X共z0,z0兲S共z0兲, 共1兲
fields upward to the surface. Rietveld 共1995兲 extends the technique
to three dimensions and demonstrates on a complex water-tank mod- with
el. Dependence of synthesis operators on the velocity model above
the target contributes to optimum illumination over the target. X共z0,z0兲 ⳱ 兺 Wⳮ共z0,zm兲R共zm兲WⳭ共zm,z0兲, 共2兲
m
The high efficiency and accuracy of areal-shot-record and con-
trolled-illumination technologies are attractive. Chen 共2004兲 pre- where Wⳮ and WⳭ are the upward and downward propagation oper-
sents a new, efficient PSDM method by surface-rotation controlled ators, respectively, between depths z0 and zm and where R共zm兲 is the
illumination based on areal shot-record technology to realize opti- reflection coefficient and X共z0,z0兲 represents the total wave-propa-
mum illumination control of the target area. The technique controls gation response. All matrices refer to one monochromatic compo-
source illumination by constructing a kernel synthesis operator nent.
through inverse wave extrapolation of a predefined source wavefield Let k and n denote the total numbers of receivers and sources, re-
at the target level. This is followed by fast angle rotation in the time spectively. Then S共z0兲 ⳱ 共sij兲k⫻n and P共z0兲 ⳱ 共pij兲k⫻n. The jth col-
domain to produce different illuminating synthesis operators 共Chen umn of S共z0兲 represents the source distribution of the jth survey ex-
et al., 2002, 2003, 2004兲 rather than complex wavefield extrapola- periment, and the jth column of P共z0兲 represents the jth shot record.
tions as found in the on-target control method of Rietveld and Each operator matrix in equation 2 describes a k ⫻ k square matrix.
Berkhout 共1994兲. This method efficiently controls illuminations of Equations 1 and 2 and the following mathematical derivations are
source wavefields directly at the surface in a target-oriented way. considered in the frequency domain.
The numerical test on Marmousi data obtains high-quality images of Let us define a complex synthesis operator L共z0兲 at surface z0 in
the target zone. the frequency domain:
Because rotation of the kernel operator can produce synthesis op-
erators with different illuminations, Chen 共2004兲 further develops L共z0兲 ⳱ 共f 1,f 2, . . . ,f n兲T, 共3兲
curved-wave migration technology by introducing the curved-wave with f i ⳱ f共xi,z0兲, and superscript T denoting the transposed vector.
base kernel. Curved-wave migration is conducted in the curved- Function f共x,z0兲 is a complex function. The operator is applied to the
wave data-domain characteristic of a ray parameter. It remains an ar- source to synthesize the areal source wavefield by the formula
eal shot-record migration method. First, it defines a complex vector
operator as a base kernel and then multiplies the base kernel by a Ssyn共z0兲 ⳱ S共z0兲L共z0兲. 共4兲
complex illumination-disturbance factor 共an exponential function兲
Multiplying the two sides of wave equation 1 by operator L共z0兲 and
to construct the synthesis operator used to synthesize the curved-
substituting expression 4, it becomes
wave source and curved-wave record data. The factor is related to a
ray parameter describing apparent slowness of the synthesized Psyn共z0兲 ⳱ X共z0,z0兲Ssyn共z0兲, 共5兲
curved source. Required synthesis operators with the fine illumina-
tion sampling necessary for migration can be constructed easily by with
sampling the ray parameter.
Synthesis operators for curved-wavefield synthesis can be de-
Psyn共z0兲 ⳱ P共z0兲L共z0兲. 共6兲
signed in a target-oriented manner. We present curved-wave PSDM In terms of equation 5, the generated wavefield Psyn共z0兲 is the seis-
by controlled illumination on the basis of curved-wave migration mic response arising from areal source wavefield Ssyn共z0兲. Thus, it is
technology. We adopt depth wave extrapolation combined with fast called an areal shot record. From expression 6, areal shot record
surface perturbations of illumination of the base kernel to generate Psyn共z0兲 is synthesized by applying operator L共z0兲 to the shot-record
target-oriented synthesis operators, improving the efficiency of con- matrix.
trolled illumination compared with the on-target control method, After proper surface-related preprocessing, we can assume that all
which depends only on wave extrapolation 共Rietveld and Berkhout, sources are the same, denoted by S共 ␻ 兲, and the source can be written
1994兲. The scheme controls illumination of the synthesized curved as S共z0兲 ⳱ S共 ␻ 兲I, with I being a unity matrix and ␻ denoting the an-
source wavefield directly at the surface in a target-oriented way and gular frequency. Hence, we have
can obtain high-quality images of the target zone efficiently. We use
the standard 2D Marmousi data set to test our method. Ssyn共z0兲 ⳱ S共␻ 兲L共z0兲, 共7兲
that is,

AREAL SHOT-RECORD MIGRATION L共z0兲 ⳱ Sⳮ1共␻ 兲Ssyn共z0兲. 共8兲

Field seismic data acquisition is a process of multiple experiments This proves that the synthesis operator is the spiked version of the ar-
of source excitation and receiver-array observation in space. Areal eal source wavefield at the surface.
shot-record techonology is based on the discrete WRW model of In areal shot-record technology, the design of synthesis operator
seismic wave propagation by Berkhout 共1985兲, represented with ma- L共z0兲 is important. It can be target- or surface-oriented. Let f j
trices, after surface-related preprocessing. According to the forward ⳱ a jeⳮi␻ ␶ j for j ⳱ 1,2, . . . ,n. Then complex synthesis operator L共z0兲
model, point-source wavelets excited at different positions at sur- can be written as
face z0 in multiple experiments form a source matrix, denoted by
L共z0兲 ⳱ 共a1eⳮi␻ ␶ 1,a2eⳮi␻ ␶ 2, . . . ,aneⳮi␻ ␶ n兲, 共9兲
S共z0兲. Accordingly, all surveyed shot records also form a record ma-
trix, denoted by P共z0兲. Shot record P共z0兲 arising from source S共z0兲 where ai is amplitude, n is the number of sources, and ␶ i represents
satisfies the wave equation the time shift at the ith source position. When, for all i ⳱ 1,2, . . . ,n,

Downloaded from http://pubs.geoscienceworld.org/geophysics/article-pdf/74/4/S95/3220115/gsgpy_74_4_S95.pdf

by CNOOC user
on 31 August 2022
 Target-oriented curved-wave PSDM S97

amplitude ai ⳱ 1 and time shift ␶ i ⳱ p共xi–xr兲, with xr being a refer- L共z0兲 ⳱ 共f 1eⳮi␻ p共x1兲x1,f 2eⳮi␻ p共x2兲x2,. . .,f neⳮi␻ p共xn兲xn兲T,
ence point and ray parameter p representing the apparent slowness
along x, the operator becomes a plane-wave synthesis operator, 共11兲
L共z0兲 ⳱ 共eⳮi␻ p共x1ⳮxr兲,eⳮi␻ p共x2ⳮxr兲, . . . ,eⳮi␻ p共xnⳮxr兲兲T, and the areal-shot in which ␶ i ⳱ p共xi兲xi is the time delay of the curved-wave-source
synthesis process is conventional plane-wave synthesis, or slant phase with respect to x over the datum of kernel f共x,z0兲, and p
stacking. ⳱ p共xi兲 is variant with x and is named the curved-wave ray parame-
For a complex subsurface, the synthesis operator should be de- ter, describing the slowness of curved-wavefield propagation rela-
signed in a target-oriented way by controlled illumination to ensure tive to the base-kernel curved wavefield. Let V denote the velocity of
proper illumination, as performed in Rietveld and Berkhout 共1994兲. the media, ␪ denote the rotated angle of the curved-wave synthesis
In this case, the synthesis operator becomes very complicated. operator L共z0兲 from the kernel, and ␸ represent the dip of the tangent
As stated above, areal source Ssyn共z0兲 and areal shot record Psyn共z0兲 of the kernel wavefront. The value p ⳱ 共2 / V兲sin共 ␪ / 2兲cos共共 ␪ / 2兲
form a pair of source and receiver wavefields satisfying wave equa- Ⳮ ␾ 兲 represents the difference between the apparent slowness
tion 5, equivalent to a seismic survey experiment. They can be mi- 共dt / dx兲 ⳱ 共sin共 ␪ Ⳮ ␾ 兲 / V兲 of the curved-wave source and the ap-
grated using conventional depth migration applicable to “point” shot parent slowness 共dt / dx兲 ⳱ 共sin ␾ / V兲 of the base kernel source
gathers. In the migration process, Ssyn共z0兲 and Psyn共z0兲 are extrapolat- along the x-direction. Base operator Lc共z0兲 corresponds to the p ⳱ 0
ed forward and inversely in the vertical direction, respectively, fol- case.
lowed by reflection imaging. A series of synthesis operators with In expression 11, each element eⳮi␻ p共x j兲x j is a factor with a spatial
finely sampled illuminations is necessary for accurate imaging. illumination disturbance p共x j兲 to function f共x,z兲, or base operator
Thus, a series of areal shot records is synthesized. For instance, in Lc共z0兲. This illumination disturbance is related to ray parameter p,
plane-wave migration, ray parameter p is sampled finely enough to the angles ␪ and ␸ . Different illumination disturbances produce dif-
produce plane-wave sources with different illuminating directions. ferent synthesis operators. The generation mechanism for the
All migration results of areal shot records are combined to create the curved-wave synthesis operator L共z0兲, or its geometric and physical
final image. meanings, is shown in Figure 1. In the figure, angle ␸ also represents
the incidence direction of the base curved-wave source deviating
from the vertical direction, and ␪ represents the incidence direction
CURVED-WAVE MIGRATION of the synthesized curved-wave source deviating from the direction
Plane waves propagate in a single direction. The wavefront 共the normal to the tangent direction of the base curved-wave source.
surface of constant phase兲 is a plane perpendicular to the propagat- According to expression 11, the excitation surface of the curved-
ing direction 共Rossi, 1988; Young, 2007兲. Here, the curved wave- wave source is a curved surface with lateral curvature variations
field is defined as a complex wavefield, with the wavefront forming a arising from the space-variant property of functions a共x,z兲 and
curved surface with space-variant curvature at any given time. Thus ␻ ␶ 共x兲 in complex function f共x,z兲, the space-variant amplitude and
it has a space-variant phase and space-variant amplitude whatever phase. The space-variant illumination disturbance ⌬p ⳱ ⌬p共xi兲 to
the time considered. The general curved wavefield means an areal ray parameter p also contributes to synthesis of the curved wavefield
wavefield. It is a simple plane wave when wavefront curvature 共nonplane wave兲.
equals a constant and a cylindrical wave when curvature along an Particularly when the base-kernel operator is a constant vector
axis direction equals a constant at any given time. Lc共z0兲 ⳱ 共1,1, . . . ,1兲T and the curved-wave ray parameter p共x兲 ⳱ p
We introduce the general curved-wave migration technique based is also a constant, synthesis operator L共z0兲 becomes
on areal shot-record technology. We reconstruct the synthesis opera-
L共z0兲 ⳱ 共eⳮi␻ px1,eⳮi␻ px2, . . . ,eⳮi␻ pxn兲T . 共12兲
tor for synthesizing a curved wavefield in the way we give the prop-
erty of space-variant disturbances of illumination directions to an ar- This case corresponds to plane-wave synthesis, where base kernel
eal source wavefield. We implement this by taking a given synthesis Lc共z0兲 is a horizontal plane-wave synthesis operator with a共x,z兲 ⳱ 1
operator as a kernel and multiplying it with a spatial illumination dis- and ␶ 共x兲 ⳱ 0, and its incidence angle is ␸ ⳱ 0. The curved-wave
turbance factor. Then we establish the curved-wave migration pro- base kernel thus reduces to a horizontal straight line, the x-axis, and
cedure. the source is excited along the horizontal datum. Hence, ray parame-

Construction of synthesis operators

Let us consider an integrable complex function f共x,z兲
⳱ a共x,z兲eⳮi␻ ␶ 共x兲 in the space-frequency domain, with a共x,z兲 and
␶ 共x兲 as real functions. We take the complex vector operator deter-
mined by f共x,z兲,

Lc共z0兲 ⳱ 共f 1,f 2, . . . ,f n兲T, 共10兲

where f i ⳱ f共xi,z0兲 is the sample of function f共x,z兲 at source location
xi at surface z0. When applied to a source to synthesize an areal Figure 1. Illustration of the curved-wave synthesis operator. Point A
source, it represents some illumination of the areal source over the is any point at the base kernel Lc共z0兲. Angle ␸ is the dip of tangent
subsurface. By taking operator Lc共z0兲 as a base kernel and giving it line lc of the kernel wavefront at point A, and angle ␪ describes the il-
spatial illumination disturbance, we construct a general formula of lumination direction of L共z0兲, normal to its tangent line l at point A,
complex vector operator L共z0兲 for curved-wave synthesis deviated from the illumination direction of kernel Lc共z0兲.

Downloaded from http://pubs.geoscienceworld.org/geophysics/article-pdf/74/4/S95/3220115/gsgpy_74_4_S95.pdf

by CNOOC user
on 31 August 2022
 S98 Chen et al.

ter p ⳱ sin ␪ / V represents the apparent slowness along the x-di- propagation operator Wⳮ 共Wapenaar and Berkhout, 1989兲. Notation
rection, the meaning of a conventional ray parameter. Lc共z0,zm兲 represents the target-oriented base synthesis operator at
Such a synthesis operator 共defined by expression 11兲 satisfies the surface z0, extrapolated from depth zm.
definition of the synthesis operator for the areal shot record in ex- The inverse of equation 13 is
pression 3. Thus, we can apply it to synthesize areal wavefields and
call them curved wavefields according to equations 4 and 6, i.e., L共zm兲 ⳱ WⳭ共zm,z0兲Lc共z0,zm兲. 共14兲
curved source Ssyn共z0兲 and curved-wave record or data Psyn共z0兲. Dis-
turbing the curved-wave ray parameter p of the synthesis operator in From equation 13, disturbing illumination of operator L共zm兲 at the
expression 11 leads to different curved-wave synthesis operators. A target level will yield a perturbation of surface operator Lc共z0,zm兲 by
series of curved-wave parameters pi共x兲 共i ⳱ 1,2, . . . 兲 should be taken the extrapolation process
to finely sample the illumination direction of the curved-wave
source for fine illumination of the subsurface, resulting in a series of ⌬Lc共z0,zm兲 ⳱ FⳭ共z0,zm兲⌬L共zm兲, 共15兲
synthesis operators. The synthesized curved wavefields Ssyn共z0兲 and
Psyn共z0兲 can be migrated using conventional depth-migration algo- which subsequently propagates back down to the operator at target
rithms suitable for shot gathers. Efficiency of curved-wave migra- level through equation 14, leading to disturbance of the source
tion is the same as in areal shot migration. wavefield at the target.
In the on-target control scheme by Rietveld and Berkhout 共1994兲,
CURVED-WAVE CONTROLLED ILLUMINATION it is necessary to finely sample illumination of operator L共zm兲 at the
target to achieve optimum imaging in the target zone. This produces
According to the controlled-illumination technology by Rietveld a series of synthesis operators with finely sampled illuminations at
and Berkhout 共1994兲, for a given target level, source wavefields are the surface by equation 13.
put on the target level. Thus, a set of synthesis operators with differ- Here we construct the desired synthesis operators with different
ent illuminations is defined at the target level. Then each individual- illuminations by disturbing illumination of the resulting base opera-
ly defined operator is extrapolated inversely upward to the surface to tor Lc共z0,zm兲, which contributes to the illumination disturbance of
produce the corresponding target-oriented synthesis operator at the operator L共zm兲 at the target, leading to a source wavefield at the tar-
surface. This is a mode of on-target source control to establish syn- get with different illumination. Base operator Lc共z0,zm兲 is a complex
thesis operators through wave extrapolations. vector and can be written in the form of expression 10:
To control illumination of the curved-wave source at the surface
in a target-oriented way, it is favorable to first determine a target-ori- Lc共z0,zm兲 ⳱ 共f m
1 2
兲 .
n T
,f m, . . . ,f m 共16兲
ented curved-wave base kernel at the surface in terms of formula 11
of the synthesis operator. Here we use the on-target source-control Here, f mi ⳱ f共xi,z0,zm兲 is the sample of function f共x,z,zm兲, which de-
mode by extrapolation to construct only the target-oriented base ker- pends on the upper geologic structure of target level zm.
nel synthesis operator. First we define an optimum source wavefield The time-space curve depicted by function f共x,z,zm兲 correspond-
at a given target and then extrapolate it inversely upward to the sur- ing to the base-kernel operator can be regarded as a new relative da-
face to produce a target-oriented synthesis operator as a curved- tum on which to excite curved-wave sources. Unlike the convention-
wave base kernel operator. Next we establish the other required syn- al horizontal datum at the surface, this relative datum is target orient-
thesis operators with different illuminations by disturbing illumina- ed and represents a curved surface related to the geologic velocity
tion of the resulting base operator according to curved-wave migra- model above the target. Hence, function f共x,z,zm兲 could be called the
tion theory, rather than using complicated extrapolation processes. target-relative base-kernel surface.
Hence we construct the general formula for velocity-constrained tar-
get-oriented surface synthesis operators by fast illumination distur-
bances of the base kernel. Then they are applied to shot records to Construction of surface synthesis operator
synthesize curved-wave records, followed by depth extrapolation As stated, illumination of subsurface targets requires fine sam-
and imaging. Each migrated curved wavefield for different illumina- pling of perturbations of the incidence directions of base kernel
tion-disturbance control is combined at common depth points. This Lc共z0,zm兲. More specifically, a series of disturbances should be made
surface-controlled method includes two means of curved-wave illu- to the curved-source wavefield excited on the target-relative base-
mination control: ray-parameter disturbance and angle rotation. kernel surface, which will generate a series of curved-wave synthe-
sis operators with different illuminations. Geometrically, this means
Construction of the kernel synthesis operator we can carry out a series of rotations around the base-kernel surface
to yield different illuminating curved waves. Because kernel opera-
First we define a desired source wavefield at a given target level tor Lc共z0,zm兲 is based on wave theory, surface synthesis operators
zm ⳱ g共x兲 by designing a synthesis operator at the target, denoted by generated by rotations of Lc共z0,zm兲 are also velocity constrained and
L共zm兲. Generally, it is designed to ensure an optimum illuminating target oriented.
source at the target, usually representing the incident direction nor- According to the general synthesis-operator expression 11, distur-
mal to the target level. After extrapolating operator L共zm兲 inversely bance to illumination directions of the curved-wave base kernel
upward to the surface, we obtain the base synthesis operator Lc共z0,zm兲 is equivalent to sampling curved-wave ray parameter p.
Because p can be expressed as a function of rotation angle ␪ , illumi-
Lc共z0,zm兲 ⳱ FⳭ共z0,zm兲L共zm兲, 共13兲
nation control can be accomplished by sampling ␪ or p. The former
where F is the inverse of the downward propagation operator WⳭ
Ⳮ is the surface-controlled illumination with angle rotation. We classi-
and can be approximated by the complex conjugate of the upward fy and describe them as follows.

Downloaded from http://pubs.geoscienceworld.org/geophysics/article-pdf/74/4/S95/3220115/gsgpy_74_4_S95.pdf

by CNOOC user
on 31 August 2022
 Target-oriented curved-wave PSDM S99

Surface-rotation controlled illumination From equation 13, as long as a source wavefield defined at target
can be extrapolated properly upward to the surface, kernel Lc共z0,zm兲
For simplicity, we consider the problem in the time domain. Tak-
will be a stable complex vector, describing a stable wavefield. Con-
ing Lc共z0,zm兲 as a kernel, we rotate it around the survey line direction
structed by continuous algorithms, i.e., angle rotation and multiply-
x to disturb the incident direction, yielding a target-oriented synthe-
ing an exponential function with ray-parameter disturbance, opera-
sis operator formulated as
tors L共t,z0,zm兲 and L共z0,zm兲 remain stable vectors. According to areal
L共t,z0,zm兲 ⳱ Lc共t Ⳮ p共␪ ,x兲x,z0,zm兲, 共17兲 shot-record synthesis theory, they can be used to multiply the two
sides of wave equation 1 to obtain equation 5. This means that the
where ␪ ⳱ ␪ 共x兲 denotes the rotation angle of L共t,z0,zm兲 from the ker- synthesized curved wavefields Ssyn共z0,zm兲 and Psyn共z0,zm兲 satisfy
nel operator, which might be a function of grid-point coordinate x. Il- wave equation 5. Therefore, such constructed synthesis operators
lumination parameter p共 ␪ ,x兲 is the curved-wave ray parameter have no special assumption for the geologic model. However, they
shown in Figure 1. Expression 17 can be transformed into the form are dependent on the model and the properly defined source at the
of expression 11 in the frequency domain by Fourier transform with target level.
respect to time. Curved wavefields Ssyn共z0,zm兲 and Psyn共z0,zm兲 are input in the mi-
gration process to be extrapolated downward in depth by using the
conventional depth-migration method applicable to shot data. The
Ray-parameter-disturbance controlled illumination procedure of synthesis and migration is repeated for synthesis opera-
Kernel operator Lc共z0,zm兲 represents the target-relative base-ker- tors produced by different disturbances of illuminating direction, ␪
nel surface with corresponding curved-wave ray parameter p共x兲 ⳱ ␪ 共x兲 or p共x兲, of the kernel synthesis operator. All migration re-
⳱ 0. Taking the relative base-kernel surface as a kernel, we disturb sults are stacked to form the total image. Figure 2 shows the flow of
ray parameter p共x兲 directly rather than rotating angle ␪ to change or the synthesis and migration process of controlled illumination.
sample the illumination directions of curved source waves. Hence, Plane-wave synthesis is characteristic of directional illumination,
we construct the surface synthesis operator in the frequency-space which decomposes a point source into each individual plane wave to
domain fetch a signal in one single illuminating direction from all shot data,
illuminating structures with similar dips. By contrast, each curved-
1 ⳮi␻ p共x1兲x1 2 ⳮi␻ p共x2兲x2 n ⳮi␻ p共xn兲xn T
L共z0,zm兲 ⳱ 共f m e ,f me , . . . ,f m e 兲 . wave synthesis operator has a space-variant illuminating direction.
共18兲 Each contains information about the upper velocity model and

The curved-wave ray parameter p共x兲 is a variable related to x in

space, which means that disturbance of the illumination direction of
the curved source wavefield might have a different size at each point.
For a constant disturbance, it follows that
1 ⳮi␻ px1 2 ⳮi␻ px2 n ⳮi␻ pxn T
L共z0,zm兲 ⳱ 共f m e ,f me , . . . ,f m e 兲 ,
共19兲
where p is invariant in space. Disturbing ray parameter is equivalent
to rotating illumination direction, but the same ray parameter might
correspond to different rotation angles at different grid points.

Synthesis and migration of curved waves

Constructed synthesis operators L共t,z0,zm兲 and L共z0,zm兲 can be
used to synthesize curved wavefields. They are dependent on the ve-
locity model above the target level because base-kernel operator
Lc共z0,zm兲 is extrapolated through the upper velocity model above the
target. When curved wavefields are extrapolated downward to the
target, this target-oriented attribute of the synthesis operators brings
up stable and simple downward-extrapolating wavefields at the tar-
get level, favorable for optimum focusing.
According to equations 4 and 6 for areal wavefield synthesis, by
applying target-oriented synthesis operators 共constructed by equa-
tions 13 and 17 or 18兲 to point source S共z0兲 and surveyed records
P共z0兲, we can synthesize the target-oriented curved source
Ssyn共z0,zm兲 and curved-wave record Psyn共z0,zm兲, respectively, as

Ssyn共z0,zm兲 ⳱ S共z0兲L共z0,zm兲 共20兲 Figure 2. Flowchart of curved-wave controlled illumination migra-

tion. The synthesize step 共top left兲 applies inputs of the yielding syn-
and thesis operators to all shot gathers to produce the required curved-
wave records. It also applies them to sources to produce areal source
Psyn共z0,zm兲 ⳱ P共z0兲L共z0,zm兲. 共21兲 wavefields as migration input.

Downloaded from http://pubs.geoscienceworld.org/geophysics/article-pdf/74/4/S95/3220115/gsgpy_74_4_S95.pdf

by CNOOC user
on 31 August 2022
 S100 Chen et al.

makes it possible to fetch illuminating signals relevant to target SYNTHETIC EXAMPLE

structures from shot data, favorable for imaging target structures.
For complex models, it is necessary to fetch or synthesize wave sig- We used the 2D Marmousi model data set generated at the Institut
nals of space-variant illumination for accurate imaging. Français du Pétrole 共IFP兲 共Versteeg, 1994兲 to test our method. The
geometry of this model is based on a profile through the north Quen-
guela trough in the Cuanza basin, with a complex reservoir at deep
depth and strong lateral velocity variations 共Figure 3兲. The synthetic
data set contains 240 shots with 96 receivers per shot; the trace inter-
val and shot interval are both 25 m, the minimum offset is 200 m, the
time sample rate is 4 ms, and the trace length is 3 s.
The first step is to give a target level. We selected a curved reflec-
tor at about 2500-m depth as the target 共see arrows in Figure 3兲,
above the reservoir structure and between an upper and lower anti-
cline. We defined the normal-incidence illuminating source wave-
field at the target level to calculate the base-kernel synthesis opera-
tor. Numerical results are shown for two controlled-illumination im-
aging approaches, i.e., by ray-parameter disturbance and by angle
rotation. In the migration process, we used the accurate Fourier fi-
nite-difference 共FFD兲 propagator 共Ristow and Ruhl, 1994兲, which
can be applied to structures with steep dip angle and strong lateral
velocity variation.
Figure 3. Marmousi velocity model. The line and arrows indicate the Let us first consider the scheme of surface controlled illumination
target and illuminations normal to the target, respectively.
by ray-parameter disturbance. We sampled 41 ray parameters p
苸 共ⳮ250,250兲 ␮s / m with constant increment dp ⳱ 12.5 ␮s / m,
a) and calculated 41 surface synthesis operators by disturbing the illu-
mination of the extrapolated kernel operator with the sampled pa-
rameters p. Figure 4a and b shows the migration results related to
coarse and fine sampling of illumination ray parameter p. Figure 4a
is the stack of images of seven curved-wave records resulting from
seven uniformly sampled ray parameters. It gives a proper image of
most structures but still presents much noise, only realizing a partial
image. Figure 4b is the final stack of images of all 41 synthesized
curved-wave records, giving a much better image of all illuminated
structures. The result proves that given a proper sampling of ray pa-
rameter, or enough illumination sampling, the method can achieve a
good image of the whole section.
For the angle rotation scheme, we set the angle increment ⌬␪
⳱ Ⳳ1°. After continuously rotating the kernel operator clockwise
and counterclockwise, respectively, with equal maximum rotating
angles for both directions, we obtained a total of 41 surface synthesis
b)
operators. Figure 5a shows the depth migration section of the 41 syn-
thesized curved-wave records. For comparison, the imaging result
of the conventional plane-wave migration 共not target oriented兲 with
FFD propagator is given also 共Figure 5b兲. It is the stack of the migra-
tions of 41 plane waves with p ranging between ⳮ350 and
350 ␮s / m.
A comparison of the migration profiles in Figures 4b and 5a shows
the same high-quality images in the deep structures and, despite illu-
mination control over the deep target, good images in the shallow
area. In the deep complex structure marked by the circle and arrows,
both figures 共obtained with surface controlled illumination兲 show
enhanced reflected energy in the area of the reservoir and the whole
anticline, more continuous events, and clearer resolution of the lay-
ers than Figure 5b 共obtained with plane-wave migration兲.
We migrated the Marmousi data using on-target controlled-illu-
Figure 4. Depth sections migrated by FFD curved-wave surface con- mination FFD migration 共Rietveld and Berkhout, 1994兲, the target
trolled illumination with ray-parameter disturbance. 共a兲 Stack of the given as above. We put 41 plane-wave source wavefields with differ-
images of seven curved-wave records corresponding to seven ray
parameters uniformly sampled between ⳮ250 and 250 ␮s / m; 共b兲 ent illumination direction at the target level, with the ray parameter
Stack of the images of 41 curved-wave records related to 41 ray pa- p 苸 共ⳮ250,250兲 ␮s / m. Figure 6 shows the final depth-migration
rameters uniformly sampled between ⳮ250 and 250 ␮s / m. section of the 41 synthesized areal shot records. The depth image

Downloaded from http://pubs.geoscienceworld.org/geophysics/article-pdf/74/4/S95/3220115/gsgpy_74_4_S95.pdf

by CNOOC user
on 31 August 2022
 Target-oriented curved-wave PSDM S101

shows improvement in the target zone compared to the plane-wave including on-target controlled illumination兲 cost nearly six times
migration result in Figure 5, and its image quality is equivalent to the less than conventional shot migration.
surface controlled illumination in Figure 4. Figure 7 shows the Figure 9 shows the responses of target-oriented synthesis opera-
depth-imaging result obtained with the FFD migration algorithm tors. Figure 9a is the kernel synthesis operator, and Figure 9b is ob-
based on the shot data set. Imaging results using curved-wave 共con- tained by disturbing the kernel operator with ray parameter p
trolled-illumination兲 migration methods 共as shown in Figures 4 and ⳱ 25 ␮s / m, corresponding to Figure 4 for surface controlled illumi-
5兲 are as accurate as the imaging result using conventional shot mi- nation. Figure 9c displays the synthesis operator, corresponding to
gration. on-target controlled illumination in Figure 6, extrapolated from the
For more detailed comparisons, we magnify the above migration curved-wave source at the target related to p ⳱ 25 ␮s / m, i.e., two
sections in the target zone marked by circles in Figures 4 and 5. Fig- sampling intervals of the ray parameter. It is similar to the operator of
ure 8a is picked up from the section of surface-rotation controlled il- the surface controlled-illumination scheme in Figure 9b. They are all
lumination in Figure 5a, Figure 8b from the plane-wave migration in complex because of the complicated velocity model yet have good
Figure 5b, and Figure 8c from the section in Figure 6 with on-target continuity with the various energy events, showing stability of the
controlled illumination. Figure 8d is taken from shot migration in algorithm generating the synthesis operators.
Figure 7. The comparison shows that migration with curved-wave In Figure 10, we give forward-propagating configurations of the
surface controlled illumination in Figure 8a gives better images with plane-wave source and controlled curved-wave sources to account
more continuous events and higher resolution in some structures. for the effects of synthesis operators on image quality. Figure 10a
This shows that the proposed curved-wave controlled illuminations shows the complicated and distorted propagating wavefield at
can achieve fine images around the complex target zone. In addition, 2000-m depth of a horizontal plane-wave source p ⳱ 0, which is not
computation with curved-wave techniques 共controlled illumination, favorable for focusing the image. Then we consider a flat target at
800-m depth with horizontal plane wave p ⳱ 0. We calculate the ker-
nel synthesis operator and extrapolate the kernel curved-wave
a)

b)
Figure 6. Depth migration section obtained from on-target con-
trolled-illumination migration scheme. Forty-one areal sources are
located at the target level, with p values ranging between ⳮ250 and
250 ␮s / m.

Figure 5. Depth sections obtained by FFD surface-rotation con-

trolled illumination and conventional plane-wave FFD migration.
共a兲 Surface-rotation controlled illumination. This is the stack of the
images of 41 curved-wave records synthesized with 41 synthesis op-
erators, resulting from 20 clockwise rotations and 20 counterclock-
wise rotations of the kernel operator with ⌬␪ ⳱ Ⳳ1°. 共b兲 Plane-
wave migration result of 41 plane-wave records with ray parameters
p ranging between ⳮ350 and 350 ␮s / m. Figure 7. FFD PSDM of conventional shot records.

Downloaded from http://pubs.geoscienceworld.org/geophysics/article-pdf/74/4/S95/3220115/gsgpy_74_4_S95.pdf

by CNOOC user
on 31 August 2022
 S102 Chen et al.

source to the target level. Figure 10b shows that the curved wavefield ward to 800- and 2000-m depth, respectively. The wavefield is well
becomes the originally defined plane wave, despite some very weak focused, very stable 共not distorted兲, and simple at 2000-m depth near
diffractions. the target. The wavefields converge with downward propagation to
Next, we take the base-kernel curved-wave source and its rotated the target zone and become simpler and more stable with good ener-
synthesis operator related to the deep target in Figure 9. Figure 10c gy focusing at the target, favorable for optimum imaging in the deep
and d shows the base-kernel curved wavefields, extrapolating down- area. Figure 10e is the extrapolated wavefield of the curved source
corresponding to the curved-wave synthesis op-
a) c) erator in Figure 9b by the curved-wave service-
controlled method. The wavefront shows a well-
focused energy distribution, stable and simple,
compared to the plane-wave source in Figure 10a,
despite some evident diffractions. Figure 10f dis-
plays the downward-extrapolated curved wave-
field at 2000 m depth, corresponding to the
curved-wave synthesis operator in Figure 9c with
on-target controlled illumination. The wavefront
has a well-focused energy contribution and only
b) d) weak diffractions.

DISCUSSION
Construction of synthesis operators is key to
the synthesis process and to the image quality of
synthesized curved or areal wavefields. In con-
trolled illumination, target-oriented synthesis op-
erators depend on the model as well as the target.
Therefore, image quality depends on four impor-
Figure 8. Comparison of migrations in the target zone for different migration methods: tant factors: synthesis-operator design, sampling
共a兲 surface-rotation controlled illumination, 共b兲 plane-wave migration, 共c兲 on-target con- of illumination parameters 共p or ␪ 兲, target selec-
trolled illumination, and 共d兲 shot migration. tion, and model complexity. Generally, for a com-
plex model and a properly given target level, the
a)
extrapolated kernel synthesis operator has nonze-
ro energy distribution with space-variant phases
caused by wavefields coming from various direc-
tions in the subsurface. Given fine samplings of
source-illumination directions at each point x,
this nonzero energy distribution will contribute to
fetching the wavefield from each point source
record in the synthesis procedure.
In this case, fine sampling of illumination pa-
rameters is enough for accurate imaging. Howev-
er, for an extremely complicated model, if the
given target function is complicated in a very
b) c) complex area 共such as below a salt structure兲,
there might be such a case that the defined source
wavefield on some point of the target level can
never travel up to the surface. In this event, direct
disturbance of the kernel operator 共even if finely
sampled兲 might miss some effective shot-record
signals if the resulting kernel synthesis operator is
absent in some areas. A plane-wave or curved-
wave source can be added in the vacant region of
the kernel operator, followed by illumination dis-
turbance to generate a series of fine illuminations.
The resulting synthesis operators are also veloci-
Figure 9. Responses of synthesis operators: 共a兲 kernel synthesis operator, 共b兲 synthesis ty-constrained and thus target-oriented, favorable
operator generated by disturbing the kernel operator with ray parameter p ⳱ 25 ␮s / m, for optimum controlled-illumination imaging of
corresponding to Figure 4 for the surface controlled illumination, and 共c兲 synthesis opera-
tor corresponding to the on-target controlled illumination in Figure 6, extrapolated from the target zone.
the curved-wave source at the target related to p ⳱ 25 ␮s / m, i.e., two sampling intervals If diffraction waves exist in the extrapolated
of the ray parameter. kernel synthesis operator, disturbances of its illu-

Downloaded from http://pubs.geoscienceworld.org/geophysics/article-pdf/74/4/S95/3220115/gsgpy_74_4_S95.pdf

by CNOOC user
on 31 August 2022
 Target-oriented curved-wave PSDM S103

a) d) given target. Then we construct model-constraint

synthesis operators with different illuminations
by simple rotations or disturbances of the base-
kernel operator rather than by complicated wave
extrapolations. In this way, our technique effi-
ciently controls source illumination of the subsur-
face target directly at the surface. This is faster
than, yet equivalent to, controlling the source
wavefield at the target level. Resulting synthesis
operators are then applied to shot records to syn-
thesize curved-wave records, followed by depth
migration.
The synthesis procedure reduces the amount of
b) e) prestack data and increases migration efficien-
cy. The target-oriented synthesis operators con-
strained by a velocity model are favorable for bet-
ter image focus and amplitude preservation in the
target zone. Therefore, curved-wave controlled
illumination migration is a special areal shot-
record migration and can obtain high-quality im-
ages of the target zone efficiently. Compared with
conventional shot migration, efficiency depends
mainly on the sample number of the curved-wave
illumination parameter 共p or ␪ 兲 rather than the
number of shot records. The efficiency increase
c) f)
is proportional to the number of shot records
divided by the sample number of the illumina-
tion parameter. Illustrations of forward-propagat-
ing wavefields for plane-wave and curved-wave
sources prove the advances of the surface and on-
target controlled-illumination schemes.
Good numerical results of 2D standard Mar-
mousi data demonstrate the effectiveness of the
method. It can be extended to three dimensions
and is promising for efficiency and accuracy in
large real-data cases, especially in three dimen-
Figure 10. Downward-propagating configurations of plane-wave and curved-wave sions.
sources. 共a兲 Complicated wavefield at 2000-m depth from plane-wave source p ⳱ 0 at the
surface. 共b兲 Wavefield extrapolating downward to the flat target layer at 800 m with the
predefined horizontal plane-wave source p ⳱ 0. 共c, d兲 Base-kernel curved wavefields, ACKNOWLEDGMENTS
corresponding to the deep curved target level, extrapolated downward to 800- and 2000
-m depth, respectively. 共e兲 Curved source extrapolated at 2000-m depth, corresponding This research was supported partially by the
to the curved-wave synthesis operator in Figure 9b by the curved-wave surface-con- Doctoral Fund of the Ministry of Education of
trolled method. 共f兲 Curved source extrapolated at 2000-m depth, corresponding to the China 共no. 200802471102兲, the National Natural
curved-wave synthesis operator in Figure 9c by on-target controlled illumination. Science Foundation of China 共Grant no.
40674062兲, the Shanghai Leading Academic Dis-
mination might cause unreal diffraction waves at the sources. Dur- cipline Project 共project B302兲, and the Joint Laboratory for Seismic
ing the summation of all migrated curved-wave records, unreal dif- Imaging, established by the School of Ocean and Earth Science at
fraction waves not contributing to real images suppress each other. A Tongji University, China, and the Shengli Geophysical Research In-
series of fully sampled illuminating synthesis operators leads to the stitute of Sinopec. The authors are grateful to Walter Rietveld and
optimum controlled illumination, although sparsely sampled illumi- Jan Thorbecke from Delft University of Technology and to three
nations can cause aliasing. This is also a general problem for areal anonymous reviewers for their careful, critical reviews and valuable
shot-record migration. suggestions. Sincere thanks also go to the editorial team for their
constructive and valuable comments.
CONCLUSIONS
Based on curved-wave migration theory, we propose an efficient REFERENCES
and accurate new PSDM method, curved-wave controlled-illumina-
tion migration, to image the subsurface in a target-oriented manner. Berkhout, A. J., 1985, Seismic migration: Imaging of acoustic energy by
wavefield extrapolation, 3rd ed.: Elsevier Science.
First, we determine a base-kernel synthesis operator for the curved ——–, 1992, Areal shot-record technology: Journal of Seismic Exploration,
wave by extrapolating inversely a predefined source wavefield at a 1, 251–264.

Downloaded from http://pubs.geoscienceworld.org/geophysics/article-pdf/74/4/S95/3220115/gsgpy_74_4_S95.pdf

by CNOOC user
on 31 August 2022
 S104 Chen et al.

Biondi, B., and G. Palacharla, 1996, 3-D prestack migration of common-azi- Rossi, M., 1988, Acoustics and electroacoustics: Artech House.
muth data: Geophysics, 61, 1822–1832. Rietveld, W. E. A., 1995, Controlled illumination in prestack seismic migra-
Chen, X. M., 2004, Prestack depth migration by target-oriented surface con- tion: Ph.D. dissertation, Delft University.
trolled illumination: Ph.D. dissertation, Tongji University 共in Chinese兲. Rietveld, W. E. A., and A. J. Berkhout, 1994, Prestack depth migration by
Chen, X. M., H. Z. Wang, J. B. Chen, and Z. T. Ma, 2002, Prestack depth mi- means of controlled illumination: Geophysics, 59, 801–809.
gration by surface controlled illumination in a target-oriented way: 72nd Rietveld, W. E. A., A. J. Berkhout, and C. P. A. Wapenaar, 1992, Optimum
Annual International Meeting, SEG, Expanded Abstracts, 1180–1183. seismic illumination of hydrocarbon reservoirs: Geophysics, 57, 1334–
Chen, X. M., H. Z. Wang, J. B. Chen, S. C. Chen, Z. T. Ma, and Z. B. Hu, 1345.
2004, Prestack depth migration by surface rotation controlled illumina- Ristow, D., and T. Ruhl, 1994, Fourier finite-difference migration: Geophys-
tion: Chinese Journal of Geophysics, 47, 342–349. ics, 59, 1882–1893.
Chen, X. M., H. Z. Wang, and Z. T. Ma, 2003, Prestack depth migration by Versteeg, R., 1994, The Marmousi experience: Velocity model determination
surface controlled illumination of multi-geologic targets: Oil Geophysical on a synthetic complex data set: The Leading Edge, 13, 927–936.
Prospecting, 38, 507–511 共in Chinese兲. Wapenaar, C. P. A., and A. J. Berkhout, 1989, Elastic wavefield extrapola-
Duquet, B., and P. Lailly, 2006, Efficient 3D wave-equation migration using tion: Redatuming of single- and multi-component seismic data: Elsevier
virtual planar sources: Geophysics, 71, no. 5, S185–S197. Science.
Foster, D. J., C. C. Mosher, and S. Jin, 2002, Offset plane wave migration: Young, M., 2007, Optics and lasers, including fibers and optical waveguides,
64th Conference and Exhibition, EAGE, Extended Abstracts, B014. 5th ed.: Science Press of China. 共First published 2000 by Springer.兲.

Downloaded from http://pubs.geoscienceworld.org/geophysics/article-pdf/74/4/S95/3220115/gsgpy_74_4_S95.pdf

by CNOOC user
on 31 August 2022