Академический Документы
Профессиональный Документы
Культура Документы
org/
ABSTRACT
INTRODUCTION
Presented at the 54th Annual International SEG Meeting December 5, 1984 in Atlanta. Manuscript received by the Editor August 13, 1984;
revised manuscript received November 20,1984.
*Gulf Oil Exploration and Production Co., Bakersfield District Office, P. O. Box 1392, Bakersfield, CA 93302
:j:Department of Mathematics, University of California, Los Angeles, 405 Hilgard Avenue, Los Angeles, CA 90024.
Gulf Oil Exploration and Production Co., New Orleans District Office, P. O. Box 61590, New Orleans, LA 70161.
**Gulf Research and Development Co., Exploration Research Division, P. O. Box 37048, Houston, TX 77236.
HLoomis Laboratory of Physics, University of Illinois, 1110 West Green Street, Urbana, IL 61801.
1985 Society of Exploration Geophysicists. All rights reserved.
903
Downloaded 10/06/13 to 107.202.74.250. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/
904
Bishop et al.
surface are denoted by Zl(X), Zz(x), ... , Zp(x), ... , Zn,(x). The
slowness in the region is modeled by a function w(x, z), which
represents the reciprocal of seismic velocity at points (x, z) in
the subsurface.
For computation, it is convenient to characterize the slowness and reflector depth functions by a finite set of parameters.
Those functions are then restricted to lie in a certain finitedimensional space of functions. The intent is to make the
dimension of that space sufficiently high that functions which
are not in the space, but may more accurately describe the real
Earth, can be well approximated by functions which do lie in
the space. However, the dimension must not be so high that the
problem of inverting the available data is hopelessly indeterminate.
Consider first the slowness function w(x, z). Divide the region
of interest into a matrix of rectangular boxes having n" columns
and nz rows (Figure 1). The horizontal dimension of a box is
typically taken as four times the COP spacing, with the vertical
dimension roughly twice that. The value of w(x, z) at the center
of the box in the kth row and tth column is denoted Wk(.
Elsewhere in the box the slowness is assumed to vary in such a
fashion that the gradient of velocity within that box is constant.
The value of slowness anywhere within the box can be obtained
from an interpolation formula which depends on W kt and on
Wk-1,t, Wk+l,t, W k t - 1 ' W k t + 1 , that is, on the values of slowness at the centers of adjacent boxes. This procedure is motivated by the ray-tracing scheme. Rays are traced from source to
receiver by a shooting method which uses the entrance angle at
each box and the velocity gradient to compute the radius of
curvature of an arc and the arclength in the box. This method is
described in greater detail in Appendix A.
The reflector depth functions Zp(x) are parameterized by the
depths at which the reflector intersects vertical boundaries of
successive columns of boxes. For reflector p these depths are
I
I
I
I Wk-I.l
I
I
I
j-- - - - - - jI
I
: Wk,I_1
I
w __ -
Wk,l
Wk,l+l
:
I
----t--- - ---t-- - - - - I
Wk+I,!
_-,"-_ _~Zp.m
__--.....,..-_Z=-'P+I. m
PROBLEM FORMULATION
Downloaded 10/06/13 to 107.202.74.250. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/
denoted Zpm' 1 ~ p ~ nr , 0 ~ m ~ nx Away from box boundaries, Zp (x) is regarded as a cubic spline function which interpolates the points specified by the Zpm'
Let M be the total number of parameters which characterize
this model of the earth. From the above discussion, it is clear
that
<I>(p)
II r(p) 11 2
[td
t(p)]'[td
t(p)].
(1)
905
V<I>(p)
o.
(2)
If ~(p) is the Jacobian matrix of t(p), that is, the matrix whose
ijth element is otdopj' then equation (2) implies
~'(p)
[t(p) - t d ]
O.
(3)
= ~'(k)~(k)L\p
~'(k)r(k)
(4)
p(k)
+ L\p.
(6)
Suppose the algorithm outlined in equations (5) and (6) converges. Call the vector to which p(k) converges p*. Then p* must
be a solution of equation (3). This follows since p(k) -> p* implies
L\p -> O. So from equation (5),
~'(p*)r(p*) =
lim
~'(k)r(k) =
O.
k~oo
INVERSION ALGORITHM
_z
906
Bishop et 81.
Downloaded 10/06/13 to 107.202.74.250. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/
W(k)
w
W(k)
z
W(k)
z
W~) =
~~(k)~~k)IN)1/2.
COP
31.1
111.1
191.1
271.1
351.1
OOOOOOOOOOOOOOOOOOODOOOOOOOOOOOOOOOOOOOOOOOOOOO~OOOOOO00
IlU
lU
...
...
21l1lll.
000000000000000000000000000000000000000000000000000000OOOOOQOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOODOC
4111.
00000000000000000000000:000000000000000000000000000000000000000000000000000000000000000000000000000000000000oooe
000000000000000000000000000000000000000000000000000000OOOOOQOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO~OOOOOOODOC
~oooooooooooooooooooooooooooooooooooooooooooooooooooooOOOOOQOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOODOC
61l1lll
:J:
lll..
lU
Bill.
11111
12111
.......................
'
Hili
161l1lll
VELOCITY IN FT/SEC
~@~~0G8DDDDDDDD0~~~~
..........
_.f 1"'., _ . _.f _.f _ ._. ,nil. II " _ .UWll _ _
_II_.n_.
6000
11000
16000
FIG. 2. Simulated earth model. Flat reflectors at 5 000, 10 000, and 15 000 ft. Velocities are laterally invariant except in the top
layer.
907
The vector L\p which is the solution of equation (5) is also the
vector that minimizes the quantity
Downloaded 10/06/13 to 107.202.74.250. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/
(8)
COP
31.8
111.8
The first term of expression (8) represents a linearized approximation to II r(k+ 1) 11 2 Typically, there are many other choices of
L\p for which this first term is of comparable magnitude. It is
known that all of these vectors differ from each other by linear
combinations of those eigenvectors of 4(k)'4(k) that correspond
to small eigenvalues (Wiggins et al., 1976; Franklin, 1970). This
includes all elements in the null space of 4(k) for which the
eigenvalues are zero, as well as a number of other components
191.8
271.8
351.8
431.8
2888.
4888
...
UJ
UJ
.... 6888
zH
.......
8888.
::
UJ
Cl
18888
12888
14888
16888
COP
31.'
191.'
271.111
351.'
431.'
...
41
....
UJ
UJ
Z
H
...::....
UJ
Cl
881
18
128
14
16.88
-..~~~.@J..@J.~.~.~.QQQQqqQQE;l[;J,~~~[!,J..[!,J.~.
-12
0
12
FIG. 4. Difference in slownesses between the simulated earth model shown in Figure 2 and "gray" model for tomography.
Bishop et al.
Downloaded 10/06/13 to 107.202.74.250. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/
908
COP
271
1'1
111
351
11.I
11.I
010 0"
....
......
...
...:z:
...z
.'
L
11.I
Cl
1
12
14
lS
NEGATIVE
CHANGE, -"S/KFT
SLO~NESS
~@~~~~0G880DDD88GG0~~OOOO~
~~_11~_W~~'~"1.'~~$~~~_I~~_I~'~I",.ti~'~~""'~Y~'~11~U~
-12
12
FIG. 5. Difference between the simulated earth model and output model after five Gauss-Newton steps.
DEPTH
COP
111..
131..
271..
351.8
431
-'I'"-----r------.,.-.......
-t---r---1~_r__,.-'"t"_..;.:;...:...-__.
4900 r-....----.......
5000 +'""<+-'-.__...._..._.--_.
i
..... - -+_._...; r . . _ .. -...."T
....
-r
-~,,,---
FIG. 6. Depth to the first reflector after one, three, and five Gauss-Newton steps. "Correct" answer would be flat lines at 5 000 ft.
Downloaded 10/06/13 to 107.202.74.250. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/
(1) the
909
cop
.. 1 Q.O
FIG.
300
.. I "
.. 4QO
7. Stacked section of Line A. The five events used in the tomographic analysis are indicated by red lines.
.2QO
LINES
Downloaded 10/06/13 to 107.202.74.250. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/
3.0
1.0
!.
'a
::T
DI
ii"
!!
C)
Downloaded 10/06/13 to 107.202.74.250. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/
911
First-guess model
First iteration
Second iteration
Third iteration
Fourth iteration
46.2
6.3
5.4
5.1
5.1
FIG. 8. Contour plot of stacking velocities for line A. A stacking-velocity anomaly can be seen on the left side of the line.
'" .....
VELOCITY
I
E
....
.....
V HIGH
I
A
I
B
COP
-...._ - -- - V LOW
......
FIG. 9. A stacking-velocity anomaly caused by a high-velocity layer. (a) Raypaths at various CDPs; red indicates rays traveling through the high-velocity layer.
Layer is assumed to cause speedup < relative to rays which miss the layer. (b) Moveout curves in CDP gathers. Red curves are best-fit hyperbolas at CDPs Band
D. (c) Stacking velocity versus CDP showing velocity anomaly.
TIME
21\
Downloaded 10/06/13 to 107.202.74.250. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/
g:
!!-
!l
.g
II'
II'
!!
FIG.
200
300
.... ,"'U.M!'I""!!IUlllll!llljW'i1iUH,W.WCWWUC.w.w1J~:jj~;,;:~llwu;.;:l~l.lilJJ!.,....u.:.;.,.;,illi!Jjl
!!"'!l'I~"""!!'U''''''lIIt!IIIW'lf''''
400
..:..aW1l~:'~i~!~~:;;.~i.~ilid'i-::--'.~;'-;;';~.,.,
..... _-.,,~":;-;li;i;jil'ji-
- ItIWjW ,"","
10. Line A converted to depth using stacking velocities from Figure 8. Effect of velocity anomaly is obvious near the left edge of the line.
100
Downloaded 10/06/13 to 107.202.74.250. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/
c1
3
Col
CD
...
go
g,
<
:=I
0'
!!.
:i"
'0
;r
.g
CD~'
FIG.
,100
.20.0
,300 , 4 Q O
"~'IS"iI'~'!('i".of::iIilYi
11. Line A converted to depth using stacking velocities calculated after statics and smoothed with a long wavelength filter.
Downloaded 10/06/13 to 107.202.74.250. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/
8000
4000
!!-
"~
.-mo
......
cg
915
Downloaded 10/06/13 to 107.202.74.250. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/
CONCLUSION
ACKNOWLEDGMENTS
-12
.,.'S)
FIG.
-l.l'l -1 - U~ I.'
US
234.a
191.7
314
231.7
1.UU ,.4)15
,.U.' .1.'
3S4.8
271.7
.~.".
154.1
151.7
l',"12U.5'l'Hi.~)l
12
-jA..,
74.1
-G.I
COP
111.7
71.75
STH
474
311.7
Downloaded 10/06/13 to 107.202.74.250. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/
554.8
3517
!l
G
15".0
151.7
-'.'2'''' "0"'2';
23".8
191.1
314.0
231.7
39".8
271.1
--12
li.lelihl.5u~.1<'.H)'
FIG. 13. Velocity-depth model for line A after four Gauss-Newton steps.
12
;'.','
',',',
-'.',',
,.r,'.
74.8
-&.0
COP
',',',
111.7
71.75
5TH
"14.0
311.7
Downloaded 10/06/13 to 107.202.74.250. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/
55".8
351.1
........'"
,.
COP'
10.0...
FIG.
.~QO
.40.0.
14. Line A converted to depth using velocities from the tomographic analysis.
.20.0
Downloaded 10/06/13 to 107.202.74.250. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/
16000
12000
8000
4000
l!
!.
'a
III
io
00
CD
919
Downloaded 10/06/13 to 107.202.74.250. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/
Tomographic processing
Event
LineA
Line B
E1
E2
E3
E4
5626
7299
9757
11912
5587
7230
9713
11 814
Average
Conventional processing
Difference
Line A
Line B
+39
+69
+44
+98
+ 62.5
5480
7280
9920
11 880
5360
7080
9600
11 620
Average =
REFERENCES
Aki, K., and Lee, W. H. K., 1976, Determination of three-dimensional
velocity anomalies under a seismic array using first P arrival times
from local earthquakes; 1. A homogeneous initial model: J. Geophys. Res., 81. 4381-4399.
Anderson, D. L., and Dziewonski, A. M., 1984, Seismic tomography:
Scientific American, 251, 60-68.
Born. M., and Wolf, E., 1959, Principles of optics: Pergamon Press Inc.
Cornuelle, B. D., 1982, Acoustic tomography: Inst. of Elect. and Electron. Eng. on Geosci. and Remote Sensing, 6E-20, 326-332.
Dines, K. A., and Lytle, R. 1., 1979, Computerized geophysical tomography: Proc. of the Inst. of Elect. and Electron. Eng., 67,1065-1073.
Doherty, S. M., and C1aerbout, J. F., 1976, Structure independent
velocity estimation: Geophysics 41, 850-881.
Forsythe, G. E., Malcolm, M. A., and Moler, C. B., 1977, Computer
methods for mathematical computations: Prentice-Hall, Inc.
Franklin, 1. N., 1970, Well-posed stochastic extensions of ill-posed
linear problems: 1. Math Anal. Appl., 31, 682-716.
Gill, P. E., Murray, W. and Wright, M. H., 1981, Practical optimization: Academic Press, Inc.
Golub, G. H., and Reinsch, c., 1970, Singular value decomposition and
least-squares solutions: Numer. Math, 14,403-420.
Hawley, B. W., Zandt, G., and Smith, R. B., 1981, Simultaneous inversion for hypocenters and lateral velocity variations: an iterative
solution with a layered model: J. Geophys. Res., 86, 7073-7086.
Difference
+ 120
+200
+320
+260
+ 225
APPENDIX A
METHOD OF TRACING RAYS
Ax
(l/w i + 1
1/w i _ 1 )/(2L\x),
(A-I)
(A-2)
w(r)
1/(I/wo + A r),
(A-3)
4>
= arccos [cos
4>0 -
x/R],
(A-4)
920
Bishop et al.
and
<P = arcsin [sin <Po
+ z/R],
Downloaded 10/06/13 to 107.202.74.250. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/
(A-7)
(A-5)
and Ij1 is the angle the gradient makes with the vertical.
These equations can be used to find the position of the ray
and its angle of exit from the box. For instance, if the ray exits
from below, the depth of exit z is known, and 4> and x then
follow from equations (A-5) and (A-4), respectively. (The value
of x obtained in this way must lie within the box; if it does not,
the assumption that the ray exits from below is in error. In this
case, it is really 4> and z that must be computed with known x.)
With the raypath determined, the path length s is given by
APPENDIX B
CALCULATION OF SLOWNESS AND DEPTH DERIVATIVES
Entries in the Jacobian matrix ~ikJ are of two kinds: derivatives of traveltime with respect to slowness, and derivatives of
traveltime with respect to reflector depth. Both are computed
immediately following the computation of the traveltimes t(pikJ).
Thus, the raypaths through the model are known when the
derivatives are computed.
In the ray tracing used here, rays are continuous at box
boundaries and refract inside boxes by moving along curvilinear paths determined by the slowness gradient. However, for
simplicity the discussion will begin with an examination of the
calculation of the slowness derivatives in a medium where the
gradient is zero inside each box, and rays refract by bending at
box boundaries. Also, reflectors will consist of piece-wise linear
functions rather than cubic splines. Some discussion of the
problem of computing the derivatives in the more general case
will then follow.
From Fermat's principle, the raypath through the medium is
that path which exhibits the minimum traveltime between the
fixed source and receiver. The minimum is taken over all paths
which lie in a certain "regular neighborhood" (Born and Wolf,
1959), and it is assumed that in that neighborhood there is only
one such path. This guarantees that the traveltime is a true
minimum rather than just an extremum.
In the finite-dimensional medium considered here, the path
through a single box is simply a straight line. Moreover, any
possible path through the entire medium can be specified by
prescribing the depths and an additional finite set of independent path parameters. The latter will be denoted by the vector q
with components qn' n = 1, 2, ... , y. (For example, one specific
qn might be the distance from a box corner to the point at which
the ray enters or leaves the box, or the distance from a fixed
point on a reflector to the point of reflection.)
Now define a real functionJ; (q, p) to be the traveltime for the
ith ray along a path through the medium specified by q. From
Fermat's principle, the true raypath traveltime ti(p) is given by
ti(p)
(B-1)
o.
n = 1,2,3 ... y.
(B-2)
q* = q*(p),
n= 1
(aJ;
oqn
I ) oq: + aJ; =
q =q.
OPj
apj
oJ;.
apj
(B-5)
q:
921
Downloaded 10/06/13 to 107.202.74.250. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/
(B-6)
where the sum is taken over all boxes through which the ray
passes. Using W kt = Pi in equation (B-5) then gives
( x m-I,Zpm-1 )
ot i
-",= Skt
uW
(B-7)
lIE EML.ut'!II!NT---..:::::--....:,..--------,,_flOW
kt
Pcos e,
(B-8)
(B-9)
!J.s
= 2!J.d cos
W kl ,
ot i
"'z
=
pm
(B-IO)
APPENDIX C
DETERMINED AND UNDETERMINED QUANTITIES
Appendix C gives a proof of statements made earlier regarding the determination of depth and slowness in a model earth in
which ray paths are straight, reflectors are flat, and the gradient
of slowness in each box is zero. (Slowness in adjacent boxes
may differ; that is, this is not a restriction to laterally homogeneous models.) It is assumed that traveltime measurements for
the model earth are available from sources and receivers placed
at locations specified as necessary below. For simplicity, it is
also assumed that reflectors lie on horizontal boundaries between boxes.
That the depth of each reflector is uniquely determined by
the data can be proven as follows: consider a ray in a single
Bishop et 81.
922
:i _sec
Downloaded 10/06/13 to 107.202.74.250. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/
91
t~ - sec 2 9 2
Z;
Z;
_
-
+ L\x~
+ L\xr
(C-3)
and, consequently,
(C-4)
t(p) =
~L\p =
0,
(C-7)
(C-5)
C2.1
--..,.--
\ ~
...J
,0
EJ
2
dz
C2.2
1~
-I
I
Zp
\1II
C2.3
FIG.
FIG.
Downloaded 10/06/13 to 107.202.74.250. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/
923
I
0
0
-I
-I
0
0
0
I
0
0
0
0
0
0
0
0
0
I
0
0
-I
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
-I
0
0
0
0
0
0 0
1 -I
0
0
0
I
0
FIG.
0
0
0
-I
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
I
0
0
-I
0
1 -I
0
0
0
0
0
0
0
0 0
0 0
1 -I
I
I
I
I