Sunteți pe pagina 1din 21

Indonesian Physical Review

Normal Moveout for Far Offset in Anisotropic Media Using


The Optimized Fomel and Stovas Method
Kertanah1, M. Syamsu Rosid1#, Humbang Purba2
1 Physics Department, Faculty of Mathematics and Natural Science, Universitas Indonesia, Indonesia. E-mail#):
syamsu.rosid@ui.ac.id
2 PPTMGB, LEMIGAS Jakarta, Indonesia. E-mail: humbangp@gmail.com

ARTICLE INFO ABSTRACT


Keywords : Optimized The normal moveout method is important for estimating the
Fomel and Stovas; anisotropic parameters of the transversally isotropic media with
Normal moveout; vertical symmetry axis (VTI). We have approximated the far-offset
Anisotropy; Far offset. moveout using the optimized Fomel and Stovas method which is based
on optimization of constant coefficients for VTI media. For a given
anellipticity parameter, we have the best accuracy after the constant
coefficient is optimized. The optimized Fomel and Stovas method can
How To Cite :
handle a much larger offset and stronger parameter of anellipticity.
Furthermore, the residual moveout of optimized Fomel and Stovas has
been compared with several methods. The optimized Fomel and Stovas
method performs great superiority to the existing methods over a wide
range of offset (offset to depth ratio up to 4) and anellipticity
parameter (0-0.5). The result shows that the optimized Fomel and
Stovas is mostly better in reducing residual moveout which is
indicated by the smallest residual moveout value. This is essential
for reducing the accumulation of error especially for deeper
substructures and indicates most soluble to tackle the hockey stick
DOI :
effect which is caused by anisotropic media at far offset.

Introduction

The seismic wave propagates to subsurface and is reflected returning to earth's surface
bringing information about the conditions of the subsurface. The travel time of seismic wave
is distance function or reflected angle between 0 − 15 for near offset, mid offset
� �
around
15 − 30 , and 30 − 45 for far offset. The distribution of the angle is
� �
depending on � �
amplitude’s differences, primer velocity, secondary velocity contrast, and density [1]. But,
seismic recording at far offset often experiences problems, such as the appearance of the
hockey stick which is caused by the effect of anisotropy in the subsurface. This pattern
cannot be reduced if it still uses normal moveout (NMO) velocity derived from hyperbolic
traveltime equation. The utilizing of NMO velocity value that is not accurately will cause
residual normal moveout. It is also going to reduce the quality of seismic subsurface and
other information. Selecting NMO velocity value accurately is still resolving the problem of
distortion of frequency at far offset.
Practically, reducing the problem, a part of seismic data at far offset is muting in the
processing step. However, the important information, like lithology changes will not be
accurate enough. An anisotropic factor is completely important for information of
heterogenous lithology so that it must be included in the travel time equation. The
researchers modified and/or made new travel time into the nonhyperbolic form so that
NMO velocity and residual moveout more accurate. The measurement of the accuracy is able
to look at the smallest residual moveout value in the big offset to depth ratio (ODR).

Time – shifted hyperbolic traveltime approximation was formulated by Castle, but it is


only
accurate at thewas
approximation offset to depth
derived ratiotravel
by using less than or equal
time and to oneparameter
anellipticity (ODR ≤ 1) [2]. Another
𝜂 relationship
which are the difference in travel time between hyperbolic and nonhyperbolic. However, it
has an accuracy of less than two (ODR < 2) [3]. Ursin and Stovas [4] derived traveltime
equation for vertical transversely isotropic (VTI) medium in the fourth order that has the
same
accuracy of Taylor series in the sixth order, but its accuracy is only at ODR < 2. Fomel and
Stovas [5] proposed a general form of approximation of nonhyperbolic equations which can
be used in the various seismic media. Rosid., et.al [6] used Fomel and Stovas method [5] to
calculate residual moveout only at ODR < 2.

This paper will calculate the value of residual moveout by varying values of eta   and
also compares four travel time errors, based on the traveltime squared for vertically
transverse isotropy (VTI) layer using the hyperbolic and nonhyperbolic approximations
(Alkhalifah and
Tsvankin [3], Fomel and Stovas [5] and optimized Fomel and Stovas[7] method at ODR ≤ 4
using synthetic and real data to show the best accuracy of travel time. To see the best travel
time approximation, we can see from the value of travel time error or residual moveout. The
smallest the travel time error value indicates most precisely ����� at large offset against
�����
for zero offsets and the method is also being able to solve the problem at far offset or big
offset
to depth ratio.

Theory

Travel time Calculation by Tener and Koehler’s Method


Tener and Koehler [8] used travel time equation of wave propagation by assuming that the
wave would propagate through the shortest distance (Fermat principal) and Snell’s law, so
that resulted equation as follows:
2 2 x 2 (1)
where : t  t 

�����
Equation (1) can be developed by applicating Taylor series expansion at even order as :

t 2 c0  c1 x 2  c2 x4  ... (2)

Equation (1) is a hyperbolic form which can be derived from geometric modeling between
source and receiver. In anisotropic case, the use of formula (1) gives normal moveout
residual big enough, so that anisotropic factor must be included in this formula.
Travel time Calculation by Alkhalifah and Tsvankin’s Method
Alkhalifah and Tsvankin [3] derived travel time square formula from the equation which was
derived by Tsvankin and Thomsen as follows :
2 2 x
2
24 x (3)
t t  −
0
 0 vnmo  1  2  x 
2 2
vnmo vnmo t 2 2 2

So, moveout for P-wave can just be explained by vertical travel time and two effective
parameters, such
a combination of as normal parameter
between moveout velocity
ε dan (�� ) and anellipticity
or combination
δ,��� parameter
of ����� and �𝜂ℎ that are
(horizontal
velocity).

2 2
2
x v 2
−h v
2
nmo x 4
(4)
t t  −
0 2 2
vnmo vnmo t
 0 vnmo  vh x 
2 2 2 2

Where :
' −
vh  vnmo 1  2 ;    (5)
1  2  1  2

Travel time Calculation Through Fomel and Stovas’ Method


Fomel and Stovas [5] modified the general form of nonhyperbolic approximation for travel
time as below.
2x Ax
4 (9)
t
2
xt
2
 
0 2
v  x
2
x
2
x 
4
v 4 t02  B v2  t04  2Bt20 v2  C v4 
 
 
1  82  8 1
A  −4 , C
B , 1  2 2
1
2
The existence of nonhyperbolic part in the approximation of travel time (8) and (9) is
controlled
by parameter A. When the value A is zero ( which is implying 𝜉 = 0 or � = � ), The equation
2
(8) is hyperbolic. When both B and C are very large, equation (9) will also be reduced to
hyperbolic form.
As for the connection with the other approximations which are proposed can be seen as
follows.
− If A = 0, Fomel-Stovas approximation reduces to hyperbolic approximation.
2 2 x2
t t 
0 2 (18)
v
− By choosing parameter A = (1 – s)/2; B = s/2; C = 0 will reduce to shifted hyperbolic [2],
which is the three parameters as below.
 1 x2 (19)
t
2
 xt 11 −  t s
2

0  0 2
 s s v
− If choose parameter � = −4��; � = 1 + 2��; � = (1 + 2��) reduces the
2
approximation
proposed by Fomel and Stovas to the form proposed by Alkhalifah and Tsvankin[3].
2 2 x2 24 x (20)
t t  −
0 2
v 4  2 x2 
v t  1  2
 0 v 4 

The difference among the hyperbolic travel time method, Castle’s travel time method, travel
time calculation by Alkhalifah’s method, and Ursin and Stovas’ method lies in the
components of Fomel and Stovas method as shown in equation (9).
The five approximations are still producing residual moveout after normal moveout (NMO)
correction. Residual commonly occurs in far offset. It can be seen by the big value of offset to
depth ratio (ODR) values. Residual is remnants of seismic events which should be
perpendicular to zero offsets after NMO correction done. Yet, the seismic event still deviates
at a certain offset. The residual is shown as the figure below.

Figure 1. Residual reduction of moveout (NMO) getting bigger at far offset [9].

The residual reduction has implications for reducing the consequences of muting at far
offset. Muting at far offsets will have impacts on the lack of information obtained on
seismic data, such as lithology, fluids, and so on [9]. The smaller residual reduction
indicates the smaller
the residual value of an NMO correction or the more precise the ����� at a certain offset
of
����� (zero-offset). This relationship is shown as the smallest percentage of the
following
equation.
t zerooffset
residual
 100%
(21)
Where :
n

T
i 1
x
− T0
residual  (22)
T0

Method

Optimized Fomel dan Stovas method


Ursin and Stovas [4] proposed normalized travel time as follows.
t (23)

t0

And normalized offset.


X
x (24)
t0VNMO

Where t is whole vertical travel time defined as t 


0 0 t , V
i
i NMO is the effective normal
2
v NMOt i
moveout (NMO) velocity expressed as VNMO   t0
[10], and vNMO is an interval of

NMO velocity defined as vNMO  a0 1  2 .


Fomel and Stovas [5] proposed a general form of moveout method which is called general
moveout approximation (GMA). In vertically transverse isotropy (VTI) medium, the
normalized Fomel and Stovas can be written as bellows.
4
4
x (25)
  
2 2
x  1  x −

1  8  8 x 
1
2 2

2 1  8  8 2x 
2

x
4
1  2
1
 1  2
1  2
2
Where  is annellipticity expressed as  
 −   [3]. For a multi-layer VTI
1  
2
model, the anisotropic parameter  is replaced by an effective anisotropic parameter (eff )
that is
tv 18

expressed as  eff
  4NMO
  i 0 NMO

8t v
4
[10].

0 NMO

The optimized Fomel and Stovas [7] as bellows.


A
4
x (26)
 2  x   1  x2 −

B
C  D  E
2
x2  2  C  D  E 2

1  F 
G
1  F x2  Hx
4
1  I 2
In the optimization process, the constant coefficients of Fomel and Stovas [5] are replaced by
letters A~I which is extended coefficients already to be optimized for varying values of  an
ODR. The optimized Fomel and Stovas can be reduced to the current form [5] when
2A=D=E=4F=8 dan C=1, while optimized Fomel and Stovas method has a number of
constant coefficients (A~I) and allows them to vary freely within perturbation around 50%-
100% on the unoptimized ones. The optimization process of constant coefficients is done by
using simulated annealing (SA) algorithm. After many numerical experiments, the constant
coefficients are obtained as shown in table 1.

Table 1. The value of constant coefficient before and after optimization

Coefficients Be
A
B
C
D
E
F
G
H
I

The objective function for optimized Fomel and Stovas method using simulated annealing
(SA) algorithm as follows.
0 0.5
 −e (27)
E  A ~ I   max a 
SA

O/D4 
e

Where  a is normalized travel time obtained from Fomel and Stovas method [5] and  e is
normalized exact travel time. The interval values of  and O / D or ODR just indicate the
limitation of value used for modeling, while  is the limitation of error.

Ray Tracing
The exact data model uses anisotropic ray tracing model.

S S S S S S R R R R R R Ofse
TW

Figure 2. Anisotropic ray tracing model.


The ray tracing model of qP, qS is done by partial differential equation [11].

G j  r −  n j
qP h j tan qP  qP h j tan jqP 
M M
j − n j  −
j 1 j 1
M
qSV  h j tan qS  qSV  h j tan qSV   (27)
n
M
j V
j − n j j
j 1
j 1

 
Where G  is used to indicate G 
qP 
,
qP  qS   qS  

,  V ,  V . This is a general 
j statement j j j j
for rotation problem at the anisotropic axis in all layers. The equation above, M (M≤N) is the
deepest layer which is passed through by rays, h j is the thickness of the jth layer, r expresses
distance between source and receiver (geophone). The number of upward (  )
qP qP
and downward (  ) from the segment of qP ray in the jth layer is given by n and n .
Generally, j j
the travel time of the ray described is according to the equation above () as follows.
qP qP 

M n M n h
h
 j   r −   j 
j  j
 qP 
j qP
− −
      
  

cos j
qP V j
qP j 1 cos j V j
j 1
qSV  qSV  
M n M n h
h
j j j j
− (28)

j 1 cos j
qS  
V
 
V
qSV  
j 1 cos j
qSV  

V
 qSV  

j j

By the same way in equation  


(),  j is used to indicate that


 j    j  j , j , j V V

qP  qP  qS   qS   . In 2D transverse isotropy (IT) ray tracing, the right

 ,
side of equation () is specifically for problem which indicates a segment of upward and
downward ray for the same mode in the same layer, commonly, that is not appropriate to the
same displacement at the horizontal axis, probably caused by any rotation in the axis of
anisotropic layers.

Result and Discussion

Synthetic Data Analysis

The initial model of the VTI single layer was made using synthetic data. The model consists
of two layers which have VP 0  2000m / s, depth (h) = 1000 m, offset (x) = 4000 m,
anisotropic
parameter   0.5, and ODR ≤ 4. The value of ODR is obtained from the comparison

between
offset and depth (x/h). The initial model can be seen in figure 3.

Vp1 = 2000 m/s


Rho1 = 2.0 gr/cc
Vp2 = 2100 m/s
Rho2 = 2.1 gr/cc
Figure 3. Single layer model.
Using anisotropic ray tracing method [11] as an exact model, we do comparison traveltime
approximation by using hyperbolic, Alkhalifah [3], Fomel and Stovas [5], and optimized
Fomel dan Stovas [7] to show the accuracy of them to exact travel time model at far offset.

Figure 4. 1D exact moveout model on an anisotropic medium

Figure 4 shows that hyperbolic method and Alkhalifah method are not good enough at far
offset. Based on the figure, hyperbolic and Alkhalifah method coincide with exact travel time
when the offset is small, but they start deviating when the offset is around 1000 m and 1500 m
respectively. Whereas, the Fomel and Stovas and optimized Fomel and Stovas method always
coincide with exact travel time from 0 m to 4000 m. It shows that Fomel and Stovas and
optimized Fomel and Stovas has good accuracy.

In figure 5 performs the residual moveout value for each method. The value  used is 0.5.
This is to see the residual moveout between Fomel and Stovas method (before optimization) )
and optimized Fomel and Stovas method (after the optimization process).
𝜼 = �. �

Figure 5. 1D comparison residual moveout between hyperbolic, Alkhalifah, Fomel and Stovas, and

optimized Fomel and Stovas with = 0.5.

It presents that the optimized Fomel and Stovas method has the smallest residual moveout
value which is almost 0% at 0 <ODR≤ 4. It indicates that the optimized Fomel and Stovas is
the best.

Accuracy of Method Analysis

The calculation of residual moveout value was done and was plotted to compare optimized
Fomel and Stovas with the other methods. Residual moveout calculation is done by varying
eta   values. They are 0.1, 0.2, 0.3, 0.4, and 0.5. The result describes the accuracy and
residual
moveout value obtained by each method at a far offset which is considered on the offset to
depth ratio (ODR). It indicates the anisotropic of the medium. The biggest value of ODR, the
most anisotropic of the medium.
(a)

(b)
(c)

(d)
(e)
Figure 6. 1D Comparison residual moveout between hyperbolic, Alkhalifah, Fomel and Stovas, and
optimized Fomel
varying values of �and
andStovas
�. for (a) 𝜂 = 0.1, (b) ) 𝜂 = 0.2 (c) 𝜂 = 0.3 (d) 𝜂 = 0.4, and (e) 𝜂 = 0.5 with

Figure 6 presents the residual moveout values to show the accuracy of each method. Taking
the same value  of 0.1 to 0.5 with varying values of   and, we can see that hyperbolic has
the biggest residual moveout values. It can not be used in the anisotropic medium because its
formula is not entering the calculation of anisotropic parameter yet, according to equation 1
[7], while Alkhalifah method [3] performs that its accuracy is still good enough when   0.1 ,
but its residual moveout starts improving when the value of  is bigger. Fomel and Stovas
method [5] is better than hyperbolic and Alkhalifah method, but it shows that the biggest
value of  the biggest residual moveout value, but optimized Fomel and Stovas [7] has the
smallest
residual value.
Residual Moveout

a. Synthetic Data

The value of residual moveout has been calculated and plotted for all methods in three
dimensions to show residual moveout which is influenced by varying values of  and ODR,
as shown in figure 7 below.
HYPERBOLIC ALKHALIFAH

FOMEL-STOVAS
OPTIMIZED FOMEL-STOVAS

Figure 7. Comparison between the 3D residual moveout of hyperbolic, Alkhalifah, Fomel-Stovas


before and after optimization solved by a simulated annealing algorithm. The values of  ranges from
0 to 0.5
and ODR ranges from 0 to 4 respectively.
Figure 7 presents the 3D residual moveout comparison of hyperbolic, Alkhalifah, Fomel and
Stovas, and optimized Fomel and Stovas. Hyperbolic and Alkhalifah method perform their
maximum residual moveout of 20% and 3.5% respectively. Fomel and Stovas method shows
that its maximum residual moveout is about 0.22%, while optimized Fomel and Stovas has
maximum residual moveout of 0.0009%. The optimized Fomel and Stovas method is the
smallest value of residual moveout and most accurate. This indicates that the extension of the
total number of the optimized constant coefficient is important for improving accuracy.

(a)

(b)
(c)

(d)
Figure 8. Comparison between residual moveout of hyperbolic, Alkhalifah, Fomel and Stovas and
optimized Fomel and Stovas by varying value of  ranges from 0 to 0.5 at ODR=1, ODR=2, ODR=3,
ODR=4 respectively.
In figure 8 above illustrates the residual moveout at each ODR with the varying value of  .
We can see that residual moveout of hyperbolic is significantly increasing along with
increasing of
 value at all of ODR because it is not entering of  calculation according to equation (1)
[7].
Alkhalifah has small residual moveout at the  value ranges from 0 to 1.5, but it starts
increasing when the  value is more than 1.5 at ODR ≤ 1. However, at ODR ≤ 2, 3, and 4,
respectively, Alkhalifah method is more significantly increasing of residual moveout along
with increasing of ODR and  value. Fomel and Stovas method has residual moveout better
and more accurate than the previous ones. Its residual is smaller than the hyperbolic and
Alkhalifah method. But, optimized Fomel and Stovas is the smallest residual and most
accurate at all of ODR with varying  value. Its residual moveout is almost 0%. The smallest
value of
residual moveout indicates the best residual reduction method. It shows that optimized
Fomel and Stovas is the best solution to solve problems, such as the hockey stick
phenomenon caused by an anisotropic medium at far offset.

b. Real Data

Real data is obtained from laboratory data [12]. In this case, we just take three samples of the
data, which are Africa shales, Africa sands, and Canadian carbonates as shown in table 2 to
see the comparison of residual moveout for all methods at far offset considered by offset to
depth ratio (ODR). The maximum offset is 4000 m. The depth value is depending on depth
for each lithology so that The value of ODR will be different for different lithology, as shown
in table 3.

Table 2. The value of anisotropic parameters at different lithology

Africa shales
Africa sands
Canadian carbonates
AFRICA SHALES
RESIDUAL MOVEOUT [%]

0.5
HIPERBOLIK
0.45 ALKHALIFAH
FOMEL STOVAS
0.4
OPTIMIZED FOMEL STOVAS

0.35

0.3

0.25

0.2

0.15

0.1

0.05

0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
OFFSET DEPTH RATIO [X/D]
(a)
AFRICA SANDS

RESIDUAL MOVEOUT [%]


0.5 HIPERBOLIK
ALKHALIFAH
0.45 FOMEL STOVAS

0.4 OPTIMIZED FOMEL STOVAS

0.35

0.3

0.25

0.2

0.15

0.1

0.05
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
0 OFFSET DEPTH RATIO [X/D]

(b)
CANADIAN CARBONATES
HIPERBOLIK
0.5 ALKHALIFAH
RESIDUAL MOVEOUT [%]

FOMEL STOVAS
0.45
OPTIMIZED FOMEL STOVAS

0.4

0.35

0.3

0.25

0.2

0.15

0.1

0.05

0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
OFFSET DEPTH RATIO [X/D]
(c)
Figure 9. Comparison of residual moveout between hyperbolic, Alkhalifah, Fomel and Stovas, and
optimized Fomel and Stovas for different lithology.

Table 3. The residual moveout for different lithology

Lithology

Africa Shales
Africa Sands
Canadian
carbonate
Table 3 shows the comparison of residual moveout for different lithologies, such as shale,
sand, and carbonate. We can see that the residual moveout of optimized Fomel and Stovas
method is mostly the smallest, even Fomel and Stovas method has the smallest residual in
sands which is 0.00000827%. However, the optimized Fomel and Stovas method has the
smallest residual moveout overall. This is because of any extension of the total number of
optimized coefficient in equation (26) that is important for decreasing the residual moveout
values. The smallest percentage of residual moveout value indicates the more precise the
offset (zero offsets). It can be said that the optimized Fomel and Stovas method is better
reducing residual moveout than the Fomel and Stovas, Alkhalifah, and hyperbolic method.

Conclusion

According to the anisotropy study results, the conclusions which are obtained from the
results of processing using synthetic and real data as follows:
1. In synthetic data, The optimized Fomel and Stovas method has the best capability of
residual reduction for the anisotropic parameter   0.5 at the offset to depth ratio less
than or equal to 4 (ODR≤ 4).
2. The residual reduction value in real data [13] for different lithology from the optimized
Fomel and Stovas has smallest value of 0.000901% compare to Fomel and Stovas
(0.135587%), Alkhalifah (1.89428%), and hyperbolic (7.9528%) for shales. For sands, the
Fomel and Stovas method has smallest residual reduction than the others which is
0.00000827%, while optimized Fomel and Stovas, Alkhalifah, and hyperbolic method
have
0.0000498%, 0.003323%, 0.300865% respectively. In carbonates, The optimized Fomel and
Stovas is the smallest residual reduction which is 0.00019271%, whereas Fomel and
Stovas method has 0.00040565%, Alkhalifah has 0.0197386%, and hyperbolic has
0.49500708%.
3. The extension of the number of coefficient in the optimized Fomel and Stovas is essential
to reduce the residual value or to increase the accuracy. The smallest residual reduction
indicates the best accuracy of the travel time method. So, the optimized Fomel and Stovas
is the best accuracy and most soluble to tackle the problems, such as hockey stick
phenomenon at far offset or at the big offset to depth ratio (ODR).

Acknowledgment

We greatly acknowledge to Indonesia Endowment Fund for Education (LPDP) for financial
support of this research. We want to thank PPTMGB LEMIGAS for permission to do
research in this study.

References

[1] Sheriff, R. E. (2002): Encyclopedic dictionary of applied geophysics, Society


of
Exploration Geophysicists, 129-241.
[2] Castle, R. J. (1994): A theory of normal moveout, Geophysics, 59, 983-
999.
[3] Alkhalifah, T. dan Tsvankin, I. (1995): Velocity analysis for transversely isotropic
media,
Geophysics, 60, 1550-1566.
[4] Ursin, B., Stovas, A. (2006): Traveltime approximations for a layered transversely
isotropic medium, Geophysics, 71, D23-D33
[5] Fomel, S., & Stovas, A. (2010). Generalized nonhyperbolic moveout, (January 2014).
https://doi.org/10.1190/1.3334323
[6] Rosid, M. S., Wicaksan, B. G. A., Purba, H. (2018, Sept) NMO Residusal Reduction on
Medium Anisotropy in Field “X” using Fomel and Stovas Method. Paper dipresentasikan di
The 8th International Conference on Theoretical and Applied Physics (ICTAP), Medan.

[7] Song, H., Zhang, J., & Yao, Z. (2018). Global optimization of generalized nonhyperbolic
moveout approximation for long-offset normal moveout Global optimization of
generalized nonhyperbolic moveout approximation for long-offset normal moveout,
(November 2017). https://doi.org/10.1190/geo2016-0431.1
[8] Taner, M. T., and Koehler, F. (1969): Velocity spectra-digital computer.
[9] Yilmaz, O. (2001). Seismic Data Analysis Volume I. Society of Exploration Geophysics: Tulsa.

[10] Alkhalifah, T. (1997): Velocity Analysis using Nonhyperbolic Moveout in Transversely


Isotropic Media, Geophysics, 62, 1839-1854.

[11] P. F. Daley, E. S. Krebes, and L. R. Lines, “methods in transversely isotropic media,” vol.
22, no. 2001, pp. 1–21, 2010.

[12] Wang, Z. (2002): Seismic Anisotropy in Sedimentary Rocks, Part 2: Laboratory Data,
Geophysics, 67, 1423-1440.

S-ar putea să vă placă și