Al Chalabi1973

Download as pdf or txt
Download as pdf or txt
You are on page 1of 13

SERIES APPROXIMATION IN VELOCITY AND TRAVELTIME

COMPUTATIONS *

BY

M. AL-CHALABI **

ABSTRACT
AL-CHALABI, M., 1973, Series Approximation in Velocity and Traveltime Computations,
Geophysical Prospecting 21, 783-795.

A rigorous proof is presented to show that over a horizontally layered ground the rms
velocity cannot exceed the stacking velocity. The proof helps to illustrate the difference
between stacking and rms velocities in a quantitative manner. The series of Taner and
Koehler (1969) is used for the purpose. Convergence of this series is tested. Including more
terms will not necessarily improve the convergence. Although the series is rapidly conver-
gent when the spread length/depth ratio is small, strong oscillations are observed when
this ratio is high.

INTRODUCTION

Consider a ground consisting of a number of uniform horizontal layers. Let


T, be the traveltime from a shotpoint to the Nth interface and back to a
receiver at distance X from the shotpoint. T, and X are then related by the
two parametric equations

T, = 2 Ii (&/V,)/(I - p”V;)~ (1)


k-l

x = 2# I! (hk Q/(1 - p”?Q’/, (2)


n-1

Vk and 12, are the velocity and thickness of the kth layer, respectively. #(= sin
0,/V,) is the ray parameter, 0, being the angle of incidence at the ith, layer
(Slotnick 1959).
To obtain a direct time-distance relationship, Taner and Koehler (1969)
related the two equations through the infinite series
T~=C,+C,X2+C~X4+...fCjX2f-2+ . . .. (3)

* Received November 1972.


** Exploration and Production Research Division, BP Research Centre, Sunbury-on-
Thames, Middlesex, U.K.
784 M. AL-CHALABI

The coefficients Ci(i = I, 2, . . ., co) are functions only of the thicknesses


and velocities of the layers.

Putting

az = 2 ; vp3hk) (i = I, 2, 3, . ..) (4)


k=1

the first two coefficients were defined by

C, = a,” = (z ; hk/Vk)’ = T; (5)


x=1

C, = al/a2 = To/2 g h, V, = I/VF~‘,, (6)


n=1
where T, is the two-way vertical traveltime and ‘u,.~~is the rms velocity.

Most current methods of obtaining stacking velocities are based on a two-


term truncation of equation [3), i.e.,
T,2 = C, + C, X2. (7)
Because of this truncation the stacking velocity, vS, does not equal the true
YWSvelocity except where the ground is homogeneous. Effectively, the stacking
velocity corresponds to a best fit hyperbola through the true traveltime
(figure I). The best fit criterion varies according to the particular method of
DISTANCE

TRUE TRAVELTIME CURVE

Fig. I. A sketch showing traveltime curves for the Nth reflector.


SERIES APPROXIMATION IN VELOCITYANDTRAVELTIMECOMPUTATIONS 785

obtaining the stacking velocity. For example, if we assume that this criterion is
based on a least squares fit which minimizes

$(C1, C,) = ii (Cl + c, x; - Ty


i=1
then

V,z = ; = [m i Xi- ( : X;)“]/[mi T,2X,2 - Zti T; li X,“],


2 i=l 6-I J-1 $=I d-1
where m is the number of stacked traces and X, and T, are respectively the
spread length and traveltime corresponding to the ith trace.
The difference between the stacking and rms velocities depends on the discrep-
ancy between the true traveltime and the time computed by the two-term
truncation (figure I). One purpose of the present work is to give a formal
proof that if the traveltime in a horizontally stratified medium is computed
according to equation (7), using the true rms velocity, then at every X,
T, > T, (8)
and therefore V rms G v8. (9)
Relation (9) has already been established by observation and by physical
reasoning (e.g. Taner, Cook, and Neidell 1970). However, the author is not
aware of any mathematical proof of it. The main result of the present proof is
the expression giving the difference between the approximate and the true
traveltimes (summarized in equation (20) and detailed in the subsequent text
and appendices). This expression helps in illustrating the difference between
stacking and rms velocities in a quantitative manner. Parameters which
contribute to this difference are discussed in appendix 3. The expression is also
useful as a basis for the design of methods which correct for this difference. In
some recent examples from the North Sea, stacking velocities exceeding the rms
velocities by more than 2% were encountered. The reduction of such discrep-
ancies is desirable in a large number of problems.
Another purpose of this work is to test the convergence of the series of
equation (3).
PROOF OFT, 3 T,
Taner and Koehler (1969) give the following relations
1.3.5.. . . (2K-3)
41 = 1, qk: = ~- (k = 2, 3,4, . ..) (IO)
2.4.6.... (zk-2)

T; zzc < Akfi2k-2


(11)
n=1

X2 = ii B,, fizk, (12)


k=1
786

where

(13)

(14)

From equation (IO) it follows that

qk - qk+l = 4,&k ’ (154


and
qk+l = q,&k- I)/zk. (Isb)
We can show that

6
I=1
q2q2w-~/(2~--)l-q~= 0 (16)
(appendix I),
and

II6-l
k?zu-t+1/(=+~ + I)1 q2u+1= 0 (17)
(appendix 2)
We can also show that if
b+c=d+e (184
and
b>d>c (18b)
then
ab ac 3 ad a,, (19)
the equality being satisfied only when the ground is homogeneous (appendix 3).
Equations (7) and (II) may be written as

Tz = a”, + z ,y Bkl pzk

and
Ti = a; + i Akp2k-2 = a”, -I- g A,,, p2k,
n=2 k=1
therefore

T; - T; = ; ,-, B,, pzk - i Aktl p2”


k-1

(20)
SERIES APPROXIMATION IN VELOCITY AND TRAVELTIME COMPUTATIONS 787

where

(21)
We now show that

d$(K h) =.&(K 4 = a, B,, - M,,, 3 0 (22)


(k = I, 2, 3, . . .) co)

which is a sufficient condition for T, > T,.

Odd Values oj k

Let zw = k + Substituting
I. (13) and (14) in (22) we get
w--l
fk(V, 4 = WC x 42 q2w-1 %+l a2,-i+l) + 4; a, ai+,
<=1
w--l
- 2a2( x qg qzw-c+l at a2w-t+1 )-2q,q,+l a2a,a,+lp
t-1
(k = I, 3, 5, . . . . ~0).
By adding o = qi a, ai +1 -4: a, a&+, to the R.H.S. we get

J&v, 12)= [2 ii 4i a2,-i+l(qzw-i at+l al -q2w-i+l at a2)l - qk ai+, a,.


*=1

Let P be any non-negative quantity regardless of its magnitude (therefore, if


two positive quantities are each denoted by P then their sum will also be
denoted by P, i.e. P + P = P).
Since,
qzW-g = qzW-t+l + qzw-i/2(2w -i) by equation (Isa)
and
az+lal=a,a2+P by relation (rg),
therefore
q2w -i %+I a1 -q2w-i+l a, a2 = P + ai+l a, q2w-i/2(2w -4,
i.e.,

fk(V,~)=P+al[{ia,,, i-1 a2~-i+1% q2~-~/(2W--i)}--~+,q~l. (23)

In equation (23) qk ai +1 cannot exceed the summation term because

5 4t qzzd(2w - 4 = 4; by equation (16),


788 M. AL-CHALABI

and
%+la,,-i+l a a;+1 by relation (19).
Therefore

Evea Values of k
Let 2% = k. Substituting (13) and (14) in (22) we get

v
- 26 (2 qt42u-t+2atia2u-i+2
1-dfl a2ai+,
$=I
= [ (2 I=1i qta2u-i+2 (q2u-2+1 q+l 4 -q2u-i+2 ai a2)l -qi+l ai,, a,

(k = 2, 4, 6, . . . , co).
Since
q2u-i+l = qzu-i+2 + qzu-i+1/2(2u -i + I) by equation (Isa)
and
al ai+, = a2 ai + P by relation (rg),
One has

I&(VA =P+a,[(C aia2,-,+2q~q2,-,+l/(2~-~+fI)~--~+lq~+ll. (24)


I-1
In equation (24)) qz +1 ~2:+ 1 cannot exceed the summation term because

Ii qiq2u-i+1/(2f+- + I) = q;+1 by equation (17)


i-1

and
a2u-i+2 al 2 ai+, by relation (19).

Therefore
.h(V, k) 3 0 (k = 2, 4, 6, . . . , co).

CONVERGENCE PROPERTIES OFEQUATION (3)


After a slight modification, the definitions of Taner and Koehler (1969) may
be written as
C3 = (4 - al a,)/44
C, = [al(ai-a2 a4) + a3 (al a,--%/ 8~;
c, = [24a, a&Z a4 - a:) + %:(a; - a, a,) + 4d(d - a2 ~4W4~~0
SERIES APPROXIMATION IN VELOCITY AND TRAVELTIME COMPUTATIONS 789

Relation (19) shows that C3 is negative (Taner and Koehler conclude this
using the Schwartz inequality) while C, and C, are each a combination of
positive and negative terms. We have already seen that the time produced by
a two-term truncation of equation (3) is greater than the true traveltime.
Therefore, convergence of the series will improve by including the negative
third term unless
c, x4 > z(Ti- T,2).
In this case, the third term will overcorrect the truncated series and the result
will be worse than truncating at two terms. Experience has shown this to be
very rare. In practice, C, X4 is very close to Ti - T,2 so that a three-term
truncation produces usually very accurate results. Because C, can be either
positive or negative and because a three-term truncation of the series produces
times which can be higher or lower than the true traveltime, the inclusion of the
fourth term will not necessarily improve the convergence of the series. For the
same reasons, including the fifth term will not necessarily improve convergence.
A similar argument applies to the higher terms. Thus, the time obtained from
a truncated form of equation (3) does not necessarily become closer to the true
traveltime as more terms are included.
Models based on actual well data were used to test these considerations. In
the model of figure z, the top six layers have high velocities such as would be
encountered in regions of a thick permafrost. The residual, which is the dif-
ference between the true traveltime and the time computed from a truncated
form of equation (3), is plotted against the number of terms retained in the
series. The seventh and thirteenth interfaces are considered. The plot shows that
including more terms does not always improve the results. Note, however, that
the three-term truncation produces consistently small residuals. Oscillations
in the residuals are most marked when the spread length/depth ratio is high.
Sensitivity of the series to the spread length/depth ratio is more clearly dem-
onstrated in figure 3. Residuals corresponding to the bottom of the seventh
layer of the previous example are plotted against spread length. The plots
shown are for cases where the series is truncated at successive terms between 2
and IO. Generally, the increase in the residual value at large spread lengths is
more pronounced when higher terms are included. The increase in sensitivity to
spread length as more terms are included is also suggested by the form of
equation (3) where the power of X increases rapidly at successive terms.
In practice, we seldom require to deal with spread length/depth ratios in
excess of 2.0. Hence, the above considerations do not usually have serious
consequences. We tested many models including 1000 models with randomly
generated velocities, at spread length/depth ratios not exceeding 2.0. Within
the first five terms there was invariably rapid convergence with little or no
790 M. AL-CHALABI

NUMBER OF SPREAD
LAYERS ABOVE
DEPTN
LENGTH
(ft.1
INTERFACE (ft.)
,560 6000

,580 6000

4280 6000

Fig. 2. A plot of the residuals vs. the number of terms included in the series of equation (3).

BOTTOM OF 7th LAYER

DEPTH:m50ft.

Fig. 3. A plot of the residuals vs. spread length when the series of equation (3) is truncated
at terms between 2 and IO. The data correspond to layer 7 of figure 2.
SERIESAPPROXIMATION IN VELOCITYANDTRAVELTIMECOMPUTATIONS 791

oscillation. When the spread length/depth ratio did not exceed 1.0, all results
were accurate to better than 0.5%. When the spread length/depth ratio in-
creased to 2.0, none of the two-term truncation results were accurate to better
than 2% and only 55% of them were accurate to better than 3%. In 97% of
these cases, a three-term truncation was sufficient to improve the accuracy to
better than 1%.

ACKNOWLEDGEMENTS
I am very grateful to Dr. P. N. S. O’Brien for reading the manuscript and
to the Chairman and Board of Directors of The British Petroleum Company
Limited for their permission to publish this work.

APPENDIX I
DERIVATION OF EQUATION (16)
The equation
s-w
q&w = ,yl (ZW + 2j - 3)/(2W + 2j - 2) W< n (AI)

follows directly from equation (IO). Dividing all terms in equation (16) by qw
and using equation (AI) we get

~- 41 (zw-1)(2~1+1)...(4~-5)+ q2
(2zf+-)(2~+1)...(4w-77)
ZW--I 2742W + 2) . . . (4w-4) zw-2 zw(zw + 2) . . . (4w-6)
(zw- I) (2w + I) . . . (4w-zm-3)
+ . . . + &g& ---___
zw(2w + 2) . . . (4w-2~2-2)

+ . . . +$&y+qw i-1 =o.


( ) (AZ)
The mth term may also be written as
IO-S-1
4m zw-I 2w+ zj--I
-.
ZW-rn4w-zm-2 I-I,=1 273 + zj-2

Let us perform the following operations in the L.H.S. of (AZ).

I. q in the last term is reduced to the order of q in the term preceding it


using relation (rgb).
2. The last two terms are then summed up together. The resulting term
becomes the last term.
3. Operations I and 2 are repeated until the last term has been summed up
with the rth term.
Geophysical Prospecting, Vol. 21 50
792 M. AL-CHALABI

We now show by induction that, at the end of the three operations, the last
term is given by

(A4)

Put I = ze,- I and substitute in (A4)

Q,-1 = - qw-l (2w2;(;fI;2)

which is the result that would be obtained by summing up the last two terms
of (AZ).
Now put r = 1z + I and substitute in (A4). Reduce qn +1 to qn using equation
(Igb). After summing up Q, +r with the rtth term defined as in (A3) we get

Qn=-qn ( wfi z$?G) 4”--T2n--z (292-1-E).


I-1
But
2w-I (92- I) (428 - 2+2- I)
$yp$----- zzz
2w-n 2w--9%
Therefore
w-n
Qn = - qn Fn n ?.f%!._’
,=1 2w + q-2

which is the same as (A4) with s replacing Y. Hence the expression for QT(y =
W-I, w-2, . ..) I) given by equation (A4) is correct.
Putting I = I in (A4) we obtain
Ql = 0
which verifies equation (16).

APPENDIX 2
DERIVATION OF EQUATION (17)

Dividing all terms in equation (17) by qu+l and using equation (AI) we get
41 (2u+ I) (2uf3) . 1 . (4U-3) qa (2u + I) (2u+3) . . . (4u-5)
+..a+
ii (2u+ 2) (2zt+4) . . . (426-2) + 224- I (224+ 2) (2~+4) . . . (4u-4)
4m .(2zl+I)(221+3)...(4u-2~-1)+...+ ‘I%-1 =4+I+
___~ ___4u
224---m + I (221+ 2) (2~+4) . . . (4%--2~2) u+22U+2 UfI
-q4u+1= 0. @I)
SERIESAPPROXIMATIONINVELOCITYANDTRAVELTIMECOMPUTATIONS 793

The mth term may also be written as

4m 2% u-m 2u+23'-I
-I-I P32)
au--m+1 qu-2m 2zt+2j-2'
,=I

We now show by induction that, at the end of the same three operations
described in appendix I, the last term is given by

Put r = u and substitute in (B3)


54-1224+1
Qu=-qu,+
which is the result that would be obtained by summing up the last two terms
of (BI). Put Y = n + I and substitute in (B3). Reduce qn + I to qn using equa-
tion (rgb). After summing up Q, +1 with the &h term defined as in (Bz) we get
u--n

But
2ab (92- I) (426- 2n + I)
2%-I---
22+%++= 2u-nn.1 '
Therefore
u--n+1
a-1
Qn=-qn2u-n+I n ;ss
f-1
which is the same as (B3) with n replacing 1. Hence the expression for Qr(y =
u,u-I, . ..) I) given by equation (B3) is correct.

Putting Y = I in (B3) we obtain


Q1=o
which verifies equation (17).

APPENDIX 3
DERIVATION AND DISCUSSION OF RELATION (rg)

The L.H.S. of relation (rg)

ab a, = (2 Ii I/;““-” hj) (2 i! VF-” hk).


f=l x-1
794 M. AL-CHALABI

Put
f3 = zb-3, y = 2c-3, 6 = 2d-3, E = 2e--3.
Therefore

$ Ub a, = ( 2 v; hj) ( 5 vi hk)
j-1 x=1
- 2 vzp+yh; + 5’ ; (VP v + v; VP,) hihk. VI)
I=1 k-l ,=k+l j k
Similarly,
N--l N
a aa a, = : v;+, h; + zl I; (Vi Pk + v;v;) hjhk’ 0)
i=1 K-l l-r+1
Note that
J/JS+=- p+Y by equation (r8a).
i-i
Subtracting (C2) from (CI) we get

(C3)

where
Qk3 = Vf V; + I’; b$ - I/‘; pk - vj: vi. F4)
We now show that
Qk+o @=1,2,...,N-I; j=k+I> h+2> . . ., NJ, w
which is a sufficient condition for
ubac bud% (19)
Let
Y = v,/v,.
Therefore
Qkj = V; I’; (Yp + Y’) - v; v; (y” + y”).

Using equation (Isa) we get


Qkj = JT;+Y (y” + yy _ yf _ yE=~+Y-‘) (C6)
zzz .;+yY8-Yy)(Yp-8-I). cc71

When (Y - I) > o, the quantities in brackets are both positive and the
inequality in (Cg) is satisfied, When (Y - I) < o, both quantities are negative
and the inequality is also satisfied. When the ground is homogeneous, (Y - I)
SERIESAPPROXIMATIONINVELOCITYANDTRAVELTIMECOMPUTATIONS 795

- o and the equality in (Cg) is satisfied. Since Qkj > o when Vj # Vk, a homo-
geneous ground is also a necessary condition for the equality in relation (19).
It is important to note from (C6) that the magnitude of Qk, increases as
vj-v,
1 Y - 1 1 increases, i.e. as increases.
i vk I’
Relation (Cg) can also be verified by putting Y = Vj/Vk which gives
Vk-- v,
Thus the magnitude of Qkj increases as the velocity
IY--I’= I TF. I
contrast between the jth and the kth layer increases. A generalised form of this
contrast is given by

Therefore, the magnitude of ab a, - ad a, gets larger as the quantity

(C8)
increases.

G gives a measure of the degree of velocity heterogeneity in the ground. It


is an important quantity which also determines the difference between average
and rms velocities. The magnitude of F(V, h) in equation (20) depends on the
magnitude of a, a, - ad a, and, therefore, on the magnitude of G but is inde-
pendent of the spread geometry. Relations (rg)-(24) and those in this Appendix
show how the difference between T, and Tc, and therefore between zls and
* v rms, depends on the heterogeneity of the ground. Equation (20) also shows how
this difference varies with increasing spread length and depth. Over the same
ground, there is always an increase in Av(= II, - v,,,) with spread length, as
p gets larger. However, for the same spread geometry, increasing depth does
not always reduce Av. When there is a large increase in heterogeneity with
depth, the consequent increase in F( V, k) can outweigh the decrease in fi and
cause Av to increase. This important possibility is contrary to the generally
stated observation that the stacking velocity increasingly approaches the true
YWLS velocity as depth increases.

REFERENCES
SLOTNICK, M. M., 1959. Lessons in seismic computing, Tulsa, SEG, p. 194.
TANER, M. T., and KOEHLER, F., 1969. Velocity spectra - Digital computer derivation and
applications of velocity functions. Geophysics 34, 859-881.
TANER, M. T., COOK, E. E., and NEIDELL, N. S., 1970. Limitations of the reflection
seismic method; lessons from computer simulations. Geophysics 35, 551-573.

You might also like