SCIENCE CHINA

Earth Sciences
• RESEARCH PAPER • May 2011 Vol.54 No.5: 647–654
doi: 10.1007/s11430-010-4124-7

Jacobian matrix for the inversion of P- and S-wave velocities and

its accurate computation method
LIU FuPing1,3*, MENG XianJun2, WANG YuMei2, SHEN GuoQiang2 & YANG ChangChun3
1
Beijing Institute of Graphic Communication, Beijing 102600, China;
2
Shengli Geophysical Research Institute of China Petrochemical Corporation (SINOPEC), Dongying 257022, China;
3
Institute of Geology and Geophysics, Chinese Academy of Sciences, Beijing 100029, China

Received July 13, 2009; accepted January 21, 2010; published online November 30, 2010

The optimization inversion method based on derivatives is an important inversion technique in seismic data processing, where
the key problem is how to compute the Jacobian matrix. The computation precision of the Jacobian matrix directly influences
the success of the optimization inversion method. Currently, all AVO (Amplitude Versus Offset) inversion techniques are
based on approximate expressions of Zoeppritz equations to obtain derivatives. As a result, the computation precision and ap-
plication range of these AVO inversions are restricted undesirably. In order to improve the computation precision and to ex-
tend the application range of AVO inversions, the partial derivative equation (Jacobian matrix equation (JME) for the P- and
S-wave velocities inversion) is established with Zoeppritz equations, and the derivatives of each matrix entry with respect to P-
and S-wave velocities are derived. By solving the JME, we obtain the partial derivatives of the seismic wave reflection coeffi-
cients (RCs) with respect to P- and S-wave velocities, respectively, which are then used to invert for P- and S-wave velocities.
To better understand the behavior of the new method, we plot partial derivatives of the seismic wave reflection coefficients,
analyze the characteristics of these curves, and present new understandings for the derivatives acquired from in-depth analysis.
Because only a linear system of equations is solved in our method, the computation of Jacobian matrix is not only of high pre-
cision but also is fast and efficient. Finally, the theoretical foundation is established so that we can further study inversion
problems involving layered structures (including those with large incident angle) and can further improve computational speed
and precision.

Jacobian matrix, Zoeppritz equations, inversion of velocities, derivatives of RCs with respect to P- and S-wave veloci-
ties, large angle

Citation: Liu F P, Meng X J, Wang Y M, et al. Jacobian matrix for the inversion of P- and S-wave velocities and its accurate computation method. Sci China
Earth Sci, 2011, 54: 647–654, doi: 10.1007/s11430-010-4124-7

Determination of rock lithology and identification of fluid hot topic for geophysical research [9–20].
in reservoirs are the ultimate objectives of oil and gas ex- Since the middle 1980s, the AVO (Amplitude Versus
ploration [1–8]. Inversion of seismic data is an important Offset) [9–20] or AVA (Amplitude Versus incident Angle)
approach to improve the precision of lithology interpreta- techniques [9–20] have been stuided extensively. Their
tion. Therefore, the application of the seismic wave reflec- theoretical foundation is Zoeppritz equations [8, 9, 14],
tion coefficients to the inversion of P- and S-wave velocities which accurately describe the relations of reflected wave
or to the inversion of petrophysical characteristics remains a amplitudes to incidence angle of waves, velocity of waves,
and density of media. Zoeppritz equations indicate that the
prediction of rock lithology is possible with seismic data.
*Corresponding author (email: fupingliu60@sina.com.cn, fupingliu@bigc.edu.cn) But due to the complicated expressions of Zoeppritz equa-

© Science China Press and Springer-Verlag Berlin Heidelberg 2010 earth.scichina.com www.springerlink.com
648 Liu F P, et al. Sci China Earth Sci May (2011) Vol.54 No.5

tions, currently all AVO inversion techniques are based on be written as [7, 8]
approximate expressions of Zoeppritz equations. Accord- sin α R pp + cos β R ps − sin α ′Tpp + cos β ′Tps = − sin α , (1)
ingly, these techniques are only suitable for small incident
angles and weak reflection interfaces. However, many cos α R pp − sin β R ps + cos α ′Tpp + sin β ′Tps = cos α , (2)
problems remain that do not satisfy the approximate condi-
tions of the simplified Zoeppritz formulas. Even for the vs1 ρ vp2
cos 2β R pp − sin 2β R ps − 2 cos 2β ′Tpp
cases where the incident angle is small, the reflection inter- v p1 ρ1 v p1
face does not always satisfy the condition of weak reflection ρ 2 vs 2
for seismic waves. Consequently, the use of approximation − sin 2β ′Tps = − cos 2β , (3)
ρ1 v p1
solutions to Zoeppritz equations may lead to large errors
accumulated in the iterative procedure of inversion, and the vs21 ρ v2
computation precision and application range of AVO inver- sin 2α R pp + vs1 cos 2β R ps + 2 s 2 sin 2α ′Tpp
v p1 ρ1 v p 2
sions are undesirably restricted.
On the other hand, because the approximate formulas of ρ2 v2
− vs 2 cos 2 β ′Tps = s1 sin 2α . (4)
Zoeppritz equations are not suitable for reflection problems ρ1 v p1
with large incident angle, the inversion of seismic waves
with large incidence angle remains a difficult problem. We where Rpp, Rps, Tpp and Tps are the reflection and refraction
have encountered many large-angle inversion problems in coefficients of the P and SV waves, respectively. Zoeppritz
practical explorations [21–31], such as those identified in equations can also be expressed in matrix form as ref. [8]:
cross-well seismic techniques, in explorations in blue water, AR=B, (5)
and in explorations of weak reflection layers under over- where
thrust faults and high-speed shielding layers. It turns out
R = ⎡⎣ R pp , R ps , Tpp , Tps ⎤⎦ , B = [b1 , b2 , b3 , b4 ] ,
T T
that the inversion of layered structures in these realistic
problems is difficult with conventional seismic processing
methods. This difficulty represents a new challenge for ⎡ a11 a12 a13 a14 ⎤
seismic explorations. The key to overcoming the difficulty ⎢a a22 a23 a24 ⎥⎥
is to sufficiently use all kinds of seismic data, whereby the A = ⎢ 21 . (6)
⎢ a31 a32 a33 a34 ⎥
precision of seismic data processing and the quality of seis- ⎢ ⎥
mic imaging are improved, and the processing method will be ⎣ a41 a42 a43 a44 ⎦
more appropriate for actual geologic problems [21–32]. Thus,
In the above equation, T denotes the transpose of matrix,
developing an inversion method without simplifying Zoep-
a11 = sin α , a12 = cos β , a13 = − sin α ′, a14 = cos β ′,
pritz equations is of paramount importance to the inversion of
elastic parameters of subsurface formations. a21 = cos α , a22 = − sin β , a23 = cos α ′, a24 = sin β ′,
The optimization inversion method based on derivatives a31=cos2β, a32=−(vs1/vp1)sin2β, a33 = −( ρ 2 / ρ1 ) (v p 2 / v p1 )
has been an important inversion technique in seismic data
cos 2β ′, a34 = −( ρ 2 / ρ1 )(vs 2 / v p1 ) sin 2β ′, a41 = (vs21 / v p1 )
processing, where the computation of Jacobian matrix is the
key to solving an inverse problem. In this paper, we present sin 2α , a42 = vs1 cos 2 β , a43 = ( ρ 2 / ρ1 )(vs22 / v p 2 ) sin 2α ′,
an accurate computation method for Jacobian matrix that a44 = −( ρ 2 / ρ1 )vs 2 cos 2β ′, b1 = − sin α , b2 = cos α , b3 =
can be used to invert for P- and S-wave velocities. The new
method accurately solves Zoeppritz equations and thus − cos 2β , b4 = (vs21 / v p1 ) sin 2α . Letting the j-th pa-
avoids the difficulty encountered when using approximate rameter of vp1, vp2, vs1 and vs2 be mj, the partial derivative of
expressions of Zoeppritz equations. The theoretical founda- Zoeppritz equations with respect to mj is given by
tion is established so that we can further study inversion ∂R ∂A ∂B
problems involving layered structures (including those with A =− R+ . (7)
∂m j ∂m j ∂m j
large-angle incident waves) and improve computational
precision and speed. Eq. (7) is called Jacobian matrix equation (JME) in the in-
version of P- and S-wave velocities, where m=
1 JME for the inversion of P- and S-wave ve- ⎡⎣ v p1 , v p 2 , vs1 , vs 2 ⎤⎦ , and R is obtained by solving eq. (5).
locities
1.2 Matrix entries of JME
1.1 JME for the inversion of P- and S-wave velocities
1.2.1 Jacobian matrix entry ∂A / ∂m j for the inversion
The reflection and refraction of P-wave at an interface be-
tween two elastic media are shown in Figure 1. At the in- of P- and S-wave velocities
terface between two elastic media, Zoeppritz equations can According to Snell’s formulas [8] v p1 / sin α = vs1 / sin β =
 Liu F P, et al. Sci China Earth Sci May (2011) Vol.54 No.5 649

⎡0 0 − sin α ′ 0⎤
⎢ 0 − tan α ′ sin α ′ 0 ⎥⎥
∂A 1 ⎢0
= , (9)
∂v p 2 v p 2 ⎢ 0 0 a33 0⎥
⎢ ⎥
⎣0 0 −a43 tan 2 α ′ 0 ⎦

Similar to eq. (9), according to the relations of α, β, α ′ and β ′

to vs1 and vs2, the partial derivatives of matrix A with respect
to vs1 and vs2 can be obtained (see Appendixes C and D):

⎡ 0 − tan β sin β 0 0⎤
⎢ 0 − sin β 0 0 ⎥⎥
∂A 1 ⎢
= , (10)
∂vs1 vs1 ⎢ −4sin 2 β (2 − tan 2 β )a32 0 0⎥
⎢ ⎥
⎣ 2a41 (3a42 − 2vs1 ) 0 0⎦
Figure 1 The reflection and refraction of P-wave at an interface between
two elastic media. P and SV designate P and SV waves, respectively, α is
the incident angle of the P-wave, β is the reflection angle of the reflected ⎡0 0 0 − tan β ′ sin β ′ ⎤
SV-wave, α′ and β′ are the refraction angles of the P and SV waves, re- ⎢0 0 0 sin β ′ ⎥
spectively, vp1, vs1, vp2 and vs2 are the propagation velocities of the P and S
⎢ ⎥
⎢ ⎛ ρ vp2 ⎞ ⎥
waves in mediums 1 and 2, respectively, ρ1 and ρ2 are the densities of ∂A 1 ⎢
= 0 0 2⎜ 2 + a33 ⎟ (2 − tan 2 β ′)a34 ⎥ .
mediums 1 and 2, respectively.
∂vs 2 vs 2 ⎢ ⎜ ρ1 v p1 ⎟ ⎥
⎝ ⎠
⎢ ⎥
⎢ ⎛ ρ2 ⎞⎥
v p 2 / sin α ′ = vs 2 / sin β ′ , and the relations of trigonometric ⎢0 0 2a43 ⎜ 3a44 + 2 vs 2 ⎟ ⎥
⎣ ⎝ ρ1 ⎠⎦
functions, we have sin β = (vs1 / v p1 ) sin α , sin α ′ = (v p 2 / (11)
v p1 ) sin α , sin β ′ = (vs 2 / v p1 ) sin α .
1.2.2 Jacobian matrix entry ∂B / ∂m j for the inversion
2 2 of P- and S-wave velocities
⎛v ⎞ ⎛ vp2 ⎞
cos β = 1 − ⎜ s1 sin α ⎟ , cos α ′ = 1 − ⎜ sin α ⎟ , The matrix entries of matrix B are [8] b1 = − sin α ,
⎜ v p1 ⎟ ⎜ v p1 ⎟
⎝ ⎠ ⎝ ⎠ b2 = cos α , b3 = − cos 2β and b4 = (vs21 / v p1 ) sin 2α , re-
2 spectively. With the above parameters, we obtain the partial
⎛v ⎞ derivatives of matrix B with respect to vp1, vp2, vs1 and vs2,
cos β ′ = 1 − ⎜ s 2 sin α ⎟ ,
⎜ v p1 ⎟ given by
⎝ ⎠
∂B 1 T
= ⎡ 0 0 −4sin 2 β −b4 ⎤⎦ , (12)
cos 2β = 1 − 2sin 2 β and sin 2β = 2sin β cos β . ∂v p1 v p1 ⎣
With above relations the partial derivatives ∂A / ∂v p1
∂B 1 T
= ⎡ 0, 0, 4sin 2 β , 2b4 ⎤⎦ , (13)
can be obtained (the derivation is shown in Appendix A): ∂vs1 vs1 ⎣
∂A 1 ∂B 1
= ⋅ = [0, 0, 0, 0]T , (14)
∂v p1 v p1 ∂v p 2 v p 2
⎡ 0 tan β sin β sin α ′ tan β ′ sin β ′ ⎤
∂B 1
⎢
⎢ 0 sin β tan α ′ sin α ′ − sin β ′ ⎥
⎥
= [0, 0, 0, 0]T . (15)
⎢ ⎥
∂vs 2 vs 2
⎛ ρ vp2 ⎞
⎢ 2 (1 − a31 )
⎢
( tan 2
β − 2 ) a32 − ⎜ 3a33 + 2 2
⎜
⎟
ρ1 v p1 ⎟⎠
( tan β ′ − 2) a34 ⎥⎥
2

⎝ 2 Numerical examples of the partial derivatives

⎢ ⎥
⎢ ⎛ ρ ⎞⎥ of RCs with respect to wave velocities
⎢ − a41 2 ( vs1 − a42 ) ( tan 2
α ′ − 1) a43 −2 ⎜ a44 + 2 vs 2 ⎟ ⎥
ρ1 ⎠ ⎦
⎣ ⎝
(8) To fully understand the characteristics of the partial deriva-
tives of RCs with respect to wave velocities, we use the
Because α, β, and β ′ are independent of vp2, the partial de- elastic parameters listed in ref. [8] to calculate the partial
rivatives of the trigonometric functions of angles α and β, derivatives of RCs with respect to P- and S-wave velocities
and β ′ with respect to vp2 are equal to zero, the partial de- respectively. The elastic parameters of the two-layer me-
rivatives of matrix A with respect to vp2 is given by (the dium are as follows: vp1=3500 m/s, vp2=vp1/0.62=5645 m/s,
derivation is shown in Appendix B) vs1=vp1/1.7=2058 m/s, the frequency of the seismic wave
650 Liu F P, et al. Sci China Earth Sci May (2011) Vol.54 No.5

f=35 Hz, vs2=vp2/1.7=3320 m/s, ρ2/ρ1=1.266, and the first

critical angle αc=38.316°. The computational results are
shown in Figures 2–9, where the incident wave is P-wave.
The definitions of the symbols in these figures are
R pp ,vp1 =| ∂R pp / ∂v p1 | , R ps ,vp1 = | ∂R ps / ∂v p1 | ; R pp ,vp 2 =
| ∂R pp / ∂v p 2 |, R ps ,vp 2 = | ∂R ps / ∂v p 2 |, R pp ,vs1 = | ∂R pp /
∂vs1 |, R ps ,vs1 =| ∂R ps / ∂vs1 |, R pp ,vs 2 =| ∂R pp / ∂vs 2 | and
R ps ,vs 2 = | ∂R ps / ∂vs 2 | . respectively.
The partial derivatives (modules) of RCs with respect to
vp1 is shown in Figure 2. We see that the curve has a sharp
peak at the first critical angle (αc=38.316°), which is a sin-
gular point (if α → α c , | ∂R ps / ∂v p1 |→ ∞, | ∂R pp / ∂v p1 | Figure 2 Partial derivatives of RCs with respect to vp1.
→ ∞ ). When computing the partial derivatives of reflection
that ∂A / ∂v p1 has a zero point within range 0<α<αc (see
coefficients of seismic waves near the critical angle, special
attention has to be paid to this phenomenon. In order to fur- Figure 3(a)), and Rpp,vp1 and Rps,vp1 each has an intersection
ther understand the characteristics of these abnormal curves, point at α=63° (see Figure 3(b)).
we plot the curves in Figure 2 again (see Figure 3), where The real and imaginary parts of the partial derivatives of
the singular point (critical angle) is removed. We observe ∂R pp / ∂v p1 and ∂R ps / ∂v p1 are shown in Figure 4. We see

Figure 3 Partial derivatives of RCs with respect to vp1.

Figure 4 Real and imaginary parts of partial derivatives of RCs with respect to vp1.
 Liu F P, et al. Sci China Earth Sci May (2011) Vol.54 No.5 651

that the first critical angle is the singular point of all curves. first critical angle is a singular point of the partial deriva-
Figures 5 and 6 indicate that the partial derivatives of tives of RCs with respect to vp1 and vp2. It is also the break-
RCs with respect to vp2 and vs1 vary with the incident angle.
In Figure 6, there is no singular point on the partial deriva-
tive curves, an important difference compared to those in
Figure 5. Figure 7 shows the real and imaginary parts of the
partial derivatives. As in Figure 6, Rpp,vs1 and Rps,vs1 in Fig-
ure 2 also exhibit large difference from Rpp,vp1 and Rps,vp1.
Here, the curves have no singular points within interval
0<α<αc, and imaginary parts of ∂R pp / ∂vs1 and ∂R ps /
∂vs1 are always zero.
The partial derivatives of RCs with respect to vs2 and
their real and imaginary parts are shown in Figures 8 and 9,
respectively. We see that the real parts of ∂R pp / ∂vs 2 and
∂R ps / ∂vs 2 are almost negative within interval 0≤α≤π/2,
where the first critical angle is a breaking point of the par-
tial derivative curve.
From Figures 2 to 9, we found that at the first critical an-
gle all partial derivative curves of RCs with respect to seis-
mic wave velocities have abnormal points. Specifically, the

Figure 7 Real and imaginary parts of partial derivatives of RCs with

Figure 5 Partial derivatives of RCs with respect to vp2. respect to vs1.

Figure 6 Partial derivatives of RCs with respect to vs1. Figure 8 Partial derivatives of RCs with respect to vs2.
652 Liu F P, et al. Sci China Earth Sci May (2011) Vol.54 No.5

Figure 9 Real and imaginary parts of partial derivatives of RCs with respect to vs2.

ing point of the partial derivatives of RCs with respect to vs1 ther study. The geophysical significance for the choice is
and vs2, an obvious indicator for the critical angle. Therefore, also a problem that needs to be further explored.
this characteristic can be used to identify the critical angle The accurate computation of the Jacobian matrix over-
of reflected waves, which in turn is helpful for the study of comes the difficulty encountered when using approximate
formation lithology. expressions of Zoeppritz equations. It forms a theoretical
foundation for the analysis of similar problems so that we
can further study inversion problems with layered structures
3 Conclusions (including those with large incident angle) and improve the
computational precision and speed.
In this paper, the JME is established based on Zoeppritz
equations. By solving the JME, we have obtained an accu-
rate scheme to compute the Jacobian matrix for the inver- Appendix A Derivation of eq. (8)
sion of P- and S-wave velocities. This method is suitable for
reflection problems not only with small incident angle and
weak reflection interface but also with large incident angle With Snell’s formulas and the relations of trigonometric
(including wide-angle) and strong reflection interface. Ac- functions ( α , β , α ′ and β ′ are the function of vp1), we
cordingly, the application range of the new method is sub- have
stantially expanded compared to the aforementioned ap-
∂ sin β v 1
proximate methods without any new restriction. = − 2s1 sin α = − sin β ,
The characteristics of partial derivative of RCs with re- ∂v p1 v p1 v p1
spect to P- and S-wave velocities have been studied with 2
vs21 1 ⎛ vs1 ⎞
numerical examples. The results show that the partial de- sin 2 α ⎜⎜ sin α ⎟
∂ cos β v 3p1 v p1 ⎝ v p1 ⎟
rivative curves of RCs with respect to wave velocities have = = ⎠
an obvious abnormal point at the first critical angle. The ∂v p1 ⎛v ⎞
2 cos β
critical angle is the singular point or the breaking point of 1 − ⎜ s1 sin α ⎟
⎜ v p1 ⎟
the partial derivative curves, which is a clear indicator for ⎝ ⎠
the existence of critical angle on these curves. When using 1 sin 2 β 1
seismic reflected waves to invert for formation parameters, = = tan β sin β
v p1 cos β v p1
special attention has to be paid to this behavior.
When the incident angle is larger than the first critical ∂ sin 2 β ∂ sin β ∂ cos β
=2 cos β + 2sin β
angle (α >αc), the partial derivative of RCs with respect to ∂v p1 ∂v p1 ∂v p1
wave velocities becomes a complex number. Both the real 1 1
and the imaginary parts of partial derivatives contain a lot of = −2 sin β cos β + 2sin β tan β sin β
v p1 v p1
information of petrophysical parameters. Therefore, when
1 1 1
dealing with wide-angle reflection inversion problems,
whether we use the real part of partial derivatives or the
=−
v p1
sin 2β +
v p1
sin 2 β tan 2 β =
v p1
(
tan 2 β − 1 sin 2β ,)
imaginary part represents a new problem that requires fur-
 Liu F P, et al. Sci China Earth Sci May (2011) Vol.54 No.5 653

∂ cos 2β ∂ sin 2 β ∂ sin β Appendix B Derivation of eq. (9)

= −2 = −4sin β
∂v p1 ∂v p1 ∂v p1
, Similar to Appendix A, α, β and β ′ are independent of vp2,
1 sin 2 β
=4 sin 2 β = 4 so the partial derivatives of the trigonometric functions of
v p1 v p1
angles α, β and β ′ with respect to vp2 are all zero. Then the
∂ sin α ′ vp2 sin α ′ partial derivatives of trigonometric functions of α ′ with
= − 2 sin α = − ,
∂v p1 v p1 v p1 respect to vp2 are

v 2p 2 ∂ sin α ′ 1 1
sin 2 α = sin α = sin α ′ ,
∂ cos α ′ v
3
1 ∂v p 2 v p1 vp2
p1
= = tan α ′ sin α ′ ,
∂v p1 cos α ′ v p1 ∂ cos α ′ 1
=− tan α ′ sin α ′ ,
∂ sin 2α ′ ∂ sin α ′ ∂ cos α ′ ∂v p 2 vp2
=2 cos α ′ + 2sin α ′
∂v p1 ∂v p1 ∂v p1 ∂ sin 2α ′ ∂ sin α ′ ∂ cos α ′
=2 cos α ′ + 2sin α ′
1 ∂v p 2 ∂v p 2 ∂v p 2
= (
sin 2α ′ tan 2 α ′ − 1 , ) 1
v p1 =
vp2
( )
sin 2α ′ 1 − tan 2 α ′ ,
∂ cos 2α ′ ∂ sin 2 α ′ 4
= −2 = sin 2 α ′ .
∂v p1 ∂v p1 v p1 ∂ cos 2α ′ ∂ sin 2 α ′ 4
= −2 =− sin 2 α ′ .
∂v p 2 ∂v p 2 vp2
Similarly, we obtain
Matrix entry α43 of A with respect to vp2 is
∂ sin β ′ sin β ′ ∂ cos β ′ 1
=− , = tan β ′ sin β ′ ,
∂v p1 ∂v p1 ∂a43 ρ v2 ρ v2 1
v p1 v p1
∂v p 2
= − 2 2s 2 sin 2α ′ + 2 s 2
ρ1 v p 2 ρ1 v p 2 v p 2
(
sin 2α ′ 1 − tan 2 α ′ )
∂ sin 2 β ′ 1 ∂ cos 2β ′ 4
∂v p1
=
v p1
(
sin 2β ′ tan 2 β ′ − 1 , )
∂v p1
=
v p1
sin 2 β ′ .
1
=− a43 tan 2 α ′ .
vp2
In this case, because α is independent of vp1, the partial
derivatives of the trigonometric functions of angle α with
respect to vp1 are zero. Then the partial derivatives of matrix
Appendix C Derivation of eq. (10)
entries of A with respect to vp1 can be written as

∂a33 ρ vp2 ρ v p 2 ∂ cos 2β ′ Here, α , α ′ and β ′ are independent of vs1. The partial
= + 2 2 cos 2β ′ − 2
∂v p1 ρ1 v p1 ρ1 v p1 ∂v p1 derivatives of trigonometric functions of β with respect to
vs1 are as follow:
ρ2 v p 2 1 ⎡ ρ2 v p 2 ⎤
= 2 ⎣
⎡ cos 2β ′ − 4sin 2 β ′⎤⎦ = ⎢ −3a33 − 2 ⎥, ∂ sin β 1 ∂ cos β 1
ρ1 v p1 v p1 ⎢⎣ ρ1 v p1 ⎦⎥ = sin β , = − tan β sin β .
∂vs1 vs1 ∂vs1 vs1
∂a34 ρ 2 vs 2 ρ v ∂ sin 2 β ′ ρ 2 vs 2 The matrix entries of A with respect to vs1 can be written as
= sin 2 β ′ − 2 s 2 = sin 2 β ′
∂v p1 ρ1 v p1
2
ρ1 v p1 ∂v p1 ρ1 v 2p1
∂a32 1 v ∂ sin 2β
ρ v 1 =− sin 2β − s1
− 2 s2 sin 2β ′ ( tan 2 β ′ − 1) ∂vs1 v p1 v p1 ∂vs1
ρ1 v p1 v p1
1 1 1
=
ρ 2 vs 2 a
sin 2β ′ ( 2 − tan 2 β ′) = 34 ( tan 2 β ′ − 2 ) ; =
vs1
a32 +
vs1
(
a32 1 − tan 2 β = )
vs1
(
a32 2 − tan 2 β , )
ρ1 v p1
2
v p1
∂a42 ∂ cos 2 β 1
∂a43 ρ 2 vs22 ∂ sin 2α ′ ρ 2 vs22 1 = cos 2 β + vs1 = (3a42 − 2vs1 ).
=
∂v p1 ρ1 v p 2 ∂v p1
=
ρ1 v p 2 v p1
sin 2α ′ tan 2 α ′ − 1 ( ) ∂vs1 ∂vs1 vs1

a43
=
v p1
(
tan 2 α ′ − 1 . ) Appendix D Derivation of eq. (11)

Using above results, we can obtain eq. (8). Similar to Appendix C ( α , β and α ′ are independent of
654 Liu F P, et al. Sci China Earth Sci May (2011) Vol.54 No.5

vs2), the partial derivatives of matrix entries of A and trigo- tion. Geophys Prospect, 1996, 44: 687–717
10 Shuey R T. A simplification of the Zoeppritz-equations. Geophysics,
nometric functions of β with respect to vs2 are as follows: 1985, 50: 609–614
11 Landro M. Discrimination between pressure and fluid saturation
∂ sin β ′ 1
= sin β ′ , Changes from time-lapse seismic data. Geophysics, 2001, 66:
∂ vs 2 vs 2 836–844
12 Zhang J F, Wapenaar K. Wave field extrapolation and pre-stack
∂ cos β ′ 1 sin β ′ 2
1 depth migration in an elastic inhomogeneous media. Geophys Pros-
=− =− tan β ′ sin β ′ . pect, 2002, 50: 629–643
∂ vs 2 vs 2 cos β ′ vs 2 13 Ma Z T. DMO for P-SV converted reflection (in Chinese). Chin J
Geophys, 1996, 39: 243–250
∂a33 ρ v p 2 ∂ cos 2β ′ ρ 2 v p 2 2 14 Li J Y, Chen X H, Hao Z J, et al. A study on multiple time-lapse
=− 2 = [1 − cos 2β ′] seismic AVO inversion (in Chinese). Chin J Geophys, 2005, 48:
∂ vs 2 ρ1 v p1 ∂vs 2 ρ1 v p1 vs 2 902–908
15 Meng X J, Jiang X D, Huang H D, et al. Generalized non-linear P
2 ⎡ ρ2 v p 2 ⎤ wave and S wave velocity inversion of prestack AVA (in Chinese).
= ⎢ + a33 ⎥ , Oil Geophys Prospect, 2004, 39: 645–650
vs 2 ⎢⎣ ρ1 v p1 ⎥⎦ 16 Shou H, Liu H, Gao J H. AVO inversion based on common shot mi-
gration (in Chinese). Appl Geophys, 2006, 3: 99–104
∂a34 ρ 1 ρ v 1 17 Gu H M, Wang J Y, Jiang T, et al. Quantitative analysis of elastic

∂ vs 2
=− 2
ρ1 v p1
sin 2β ′ − 2 s 2
ρ1 v p1 vs 2
(
sin 2β ′ 1 − tan 2 β ′ ) parameter information contained in AVO of sea-bottom multi-wave
multi-component seismic data (in Chinese). Oil Geophys Prospect,
2000, 35: 56–63
a34 18 Chen J J, Yin X Y, Zhang G H. Prestack AVO inversion of layered
= ⎡ 2 − tan 2 β ′⎤⎦ . medium (in Chinese). Oil Geophy Prospect, 2006, 41: 656–662
vs 2 ⎣ 19 Wang B L, Yin X Y, Zhang F C. Gray approximation-based elastic
wave impedance equation and inversion (in Chinese). Oil Geophys
∂a44 ρ ρ ∂ Prospect, 2007, 42: 435–439
= − 2 cos 2 β ′ − 2 vs 2 cos 2 β ′ 20 Ding J C, Chang X, Liu Y K, et al. Rapid method for acoustic
∂ vs 2 ρ1 ρ1 ∂vs 2 wave-equation WTW inversion of cross-hole seismic data (in Chi-
ρ2 ρ ⎛ 4 ⎞ nese). Chin J Geophys, 2007, 50: 1527–1533
=− cos 2 β ′ − 2 vs 2 ⎜ − sin 2 β ′ ⎟ 21 Brittan J, Warner M. Wide-angle seismic velocities in heterogeneous
ρ1 ρ1 ⎝ vs 2 ⎠ crust. Geophys J Intl, 1997, 129: 269–280
22 Zhang W P, Guo P, Hu T Y. Study and practice of wide-angle seis-
1 ⎛ ρ2 ⎞ mic data processing. Appl Geophys, 2004, 1: 31–37
= ⎜ 3a44 + 2 vs 2 ⎟ . 23 Fruehn J, White R S, Fliedner M, et al. Large-aperture seismic: Im-
vs 2 ⎝ ρ1 ⎠ aging beneath high-velocity strata. World Oil, 1999, 220: 109–113
24 Liu F P, Zhang H M, Yang C C, et al. The lateral shift of quasi total
reflection of inhomogeneous P-electromagnetic wave on the interface
of conductive media (in Chinese). Chin J Geophys, 2007, 50:
This work was supported by Funding Project for Academic Human Re-
556–566
sources Development in Institutions of Higher Learning (Grant No.
25 Liu F P, Wang A L, Chen Q, et al. The research progress on lateral
PHR(20117145)) and National Natural Science Foundation of China
shift of reflected electromagnetic wave at the interface of two con-
(Grant No. 10705049).
ductive media. Chin Sci Bull, 2008, 53: 961–968
26 Hu Z P, Guan L P, Gu L X, et al. Wide angle seismic wave field
analysis and imaging method below the high velocity shield layers (in
1 Batzle M, Wang Z. Seismic properties of pore fluids. Geophysics, Chinese). Chin J Geophy, 2004, 47: 88–94
1992, 57: 1396–1408 27 Li H S. The application of wide-angle reflection method in central
2 Han D H, Nur A, Morgan D. Effects of porosity and clay content on Tarim basin (in Chinese). Geophys Prospect Petrol, 2005, 44:
wave velocities in sandstones. Geophysics, 1986, 51: 2093–2107 292–295
3 Keys R G, Xu S Y. An approximation for the Xu-White velocity 28 Wang Z, He Z H, Huang D J, et al. A seismic survey method for
model. Geophysics, 2002, 67: 1406–1414 weak reflectors below shielded layer of high velocity: Wide angle
4 Kuster G T, Toksöz M N. Velocity and attenuation of seismic waves reflection (in Chinese). Progr Explor Geophys, 2002, 25: 23–27
in two phase media: Part I: Theoretical formulation. Geophysics, 29 Sun CY, Ni CK, Li S J, et al. Feature of wide-angle reflection data
1974, 39: 587–606 and correction method (in Chinese). Oil Geophys Prospect, 2007, 42:
5 White L, Castagna J. Stochastic fluid modulus inversion. Geophysics, 24–29
2002, 67: 1853–1843 30 Wang Z, He Z H, Huang D J, et al. The wave-field features research-
6 Xu S, White R E. A new velocity model for clay-sand mixtures. ing of wide-angle reflection and the analyzing of forward modeling
Geophys Prospect, 1995, 43: 91–118 (in Chinese). Prog Geophys, 2003, 18: 116–121
7 Fred J H. Sun X P, Zhao L W, translation. Seismic Amplitude Inter- 31 Xu W J, Yu W H, Hu Z P. Feature and forward simulation of
pretation (in Chinese). Beijing: Petroleum Industry Press, 2006. wide-angle reflection (in Chinese). Oil Geophys Prospect, 2006, 41:
18–39 390–395
8 Lu J M. Seismic Exploration Principle (in Chinese). Dongying: China 32 Wang Y F. Computational Method for Inverse Problem and Their
Unversity of Petroleum Press, 1993. Vol. 1: 20–30, Vol. 2. 123–194 Applications (in Chinese). Beijing: Higher Education Press, 2007.
9 Xu S, White R E. A physical model for shear-wave velocity predic- 31–84