This article has been accepted for publication in a future issue of this journal, but has not been

fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TGRS.2021.3138750, IEEE
Transactions on Geoscience and Remote Sensing
> IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING < 1

Inversion method of elastic and fracture parameters of

shale reservoir with a set of inclined fractures
Ziyu Qin, Xiaotao Wen, Dongyong Zhou, Bo Li, Jingyi Chen

 Index Terms—monoclinic medium, shale reservoir, fracture

Abstract—The inversion of reservoir elastic parameters and weaknesses, rock physics, bayesian inversion method.
fracture parameters is of great significance to oil and gas
production. In shale reservoirs with inclined fractures, I. INTRODUCTION
using the reflection coefficient equation of vertical
fractures or horizontal fractures under the VTI (vertical
transverse isotropy) background has certain limitations.
A S the migration pathway and storage space for oil and gas,
fractures are the focus of reservoir exploration and
development research. Goodway et al. pointed out that when
For this reason, based on the linear slip model, this paper there are fractures in the underground medium, the seismic
establishes the approximate stiffness matrix of the reflection amplitude characteristics are quite different from
monoclinic medium with a set of inclined fractures under those of the medium without fractures [1]. The technique of
the background of VTI, and combines the Born scattering characterizing underground cracks based on the theory of
theory to further derive the PP wave linear reflection seismic amplitude anisotropy has achieved good application
coefficient equation of the monoclinic medium. The effects[2-5]. This is due to the weak anisotropy (WA) model
equation includes parameters such as Young's modulus, proposed by Thomsen [6], which simplifies the derivation of
Poisson's ratio, density, fracture weaknesses, and bedding seismic wave reflection/transmission coefficients under various
weaknesses. Then, a rock physics model that anisotropic media conditions. On this basis, the first-order
comprehensively considers horizontal bedding and inclined perturbation theory proposed by Jech and Pšenčík [7] has
fractures is established, and the influence of the inclination become a favorable tool for the reflection coefficient derivation
of the fractures on the reflection coefficient is analyzed. of seismic inversion in anisotropic media. Based on this, Ruger
Finally, taking the advantage of Bayesian theory, the [8] derives the PP-wave reflection coefficient of the interface
fracture weakness inversion method based on the azimuth separating the two-layer horizontal transverse isotropy (HTI),
amplitude difference is established, and the inversion vertical transverse isotropy (VTI), and orthotropic media, and
accuracy is improved by adding Cauchy constraints and analyzes the change of the reflection characteristics with the
smoothing model regularization terms. Then take the azimuth angle under these conditions. With further research,
fracture weaknesses as inputs to invert the parameters such Pšenčík and Vavryčuk [9-10] derived the PP-wave R/T
as Young's modulus, Poisson's ratio, density and bedding coefficient of the interface separating two arbitrary weakly
weaknesses. The inversion results of theoretical data show arbitrary anisotropic media. In addition to the P-wave reflection
that the method can invert the fracture information well, coefficient equation, the reflection coefficient equation of
and it can also be effectively applied to low-to-medium converted wave under different media conditions has also been
signal-to-noise ratio (SNR) data. The inversion results of derived [11-13]. Although the reflection coefficient equations
theoretical and actual data prove that the inversion in various media can be derived by simple algebraic operation,
equation derived in this paper can be used to obtain more the derivation process based on the first-order perturbation
accurate elastic and fracture parameters in fractured shale theory is still very cumbersome. Therefore, starting from the
reservoirs. first-order perturbation of density and elastic stiffness, Shaw
and Sen [14] deduced the general formula of reflection
coefficient equation under anisotropic conditions by using Born
This work is supported by the National Natural Foundation of China (grant scattering theory, and provided a simpler linearization method
no. 41774142) and in part with the National Science Foundation of China (grant of reflection coefficient. Then, many scholars have studied the
no. 42104131). (Corresponding author: Xiaotao Wen.)
Ziyu Qin is with State Key Laboratory of Oil and Gas Reservoir Geology
reflection coefficient equation under different fracture media
and Exploitation and School of Geophysics, Chengdu University of conditions, and applied the derived reflection coefficient
Technology, Chengdu 610059, China (e-mail: ziyu.qin@hotmail.com). equation to the inversion of fracture parameters in practical data.
Xiaotao Wen, Dongyong Zhou and Bo Li are with State Key Laboratory of It is generally believed that carbonate and igneous fractured
Oil and Gas Reservoir Geology and Exploitation and School of Geophysics,
Chengdu University of Technology, Chengdu 610059, China (e-mails:
reservoirs are HTI media, that is, isotropic background media
wenxiaotao@cdut.cn, zhou_dongyong@163.com, libo_sr@163.com, with a group of vertically arranged fractures[15]. Amplitude
respectively). versus incident and azimuthal angle (AVAz) inversion method
Jingyi Chen is with Seismic Anisotropy Group, Department of Geosciences, based on HTI medium has been widely studied in the past few
The University of Tulsa, Tulsa, OK 74104, USA (e-mail:
jingyi-chen@utulsa.edu).
decades and achieved good results in fracture parameter
. inversion [16-17], but this assumption has large errors in
fracture prediction of sedimentary reservoir [18-20]. In addition

1
0196-2892 (c) 2021 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: Southwest Petroleum University. Downloaded on December 30,2021 at 15:55:55 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TGRS.2021.3138750, IEEE
Transactions on Geoscience and Remote Sensing
> IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING < 2

to vertically arranged fractures, many studies have proved the containing normal and tangential fracture weakness parameters
existence of inclined fractures in underground media [21-23].  N and  T instead of Thomsen anisotropic parameters. The
This TI medium with inclined symmetry axis is defined as
definition of  N and  T are as follows:
inclined transverse isotropic (TTI) medium, and its reflection
coefficient equation can be transformed and deduced from the 4e
N  (2)
stiffness matrix of HTI or VTI medium through bond  1  k f  3 / 4  f 
transformation [24-25]. Numerical experiments show that 3 g (1  g ) 1   
ignoring the oblique inclination of the symmetry axis of TTI   g 1  g    
medium will produce great errors in determining the 16e
T  (3)
anisotropic parameters [14].  4   f 
3(3  2 g ) 1  
In fractured shale reservoirs, the anisotropy is not only
affected by fractures, but also by layered media. Therefore, this    3  2 g    
paper studies the monoclinic medium model embedded with a
where g   / M ，  f and k f are the shear modulus and bulk
group of inclined rotation invariant fractures in the background
of VTI. Using the relationship between the Hudson [26-27] modulus of the fillings in fractures respectively, when the
penny crack model and the Schoenberg [28-29] linear slip fracture is dry, k f  0 ,  f  0 .  represents aperture. e is
model, the weakness can be connected to the fracture the fracture density wich is consistent with the crack density
characteristics (fracture density and aspect ratio). Based on defined in the Hudson model [26-29]:
Shaw and Sen's method, we derive the linear expression of the N 3
reflection coefficient of the above monoclinic medium, which e  a3  (4)
includes the parameters reflecting the degree of fracture V 4
development. Using fracture-medium rock physics modeling, where, a is the radius of the crack, N/V is the number of cracks
the effect of fracture inclination angle on reflection coefficient per unit volume,  is the crack porosity,  the fracture aspect
under different fracture densities is studied, and it is proved that
ratio. In order to distinguish, the weaknesses induced by
the influence of fracture inclination angle on reflection
coefficient cannot be ignored. In order to improve the accuracy fracture is represented by  N 2 and  T2 , and the weaknesses
of the inversion, the amplitude difference between the two corresponding to bedding is represented by  N1 and  T1 .
azimuth seismic data is used to eliminate the influence of the x
VTI background on the reflection coefficient, and combined Under the assumption of small weaknesses, CVTI and S f 1
with Bayesian theory, the fracture weaknesses are first inverted. are inserted into equation (1), and the term containing
Using the PYD formula [30] as the isotropic part, the Poisson's  N1  N 2 ,  T1  T2 ,  N1  T2 ,  T1  N 2 are ignored to obtain the
ratio and Young's modulus of the reservoir can be further
inverted. By adding different noises to the synthetic data, the approximate stiffness matrix:
stability and accuracy of the algorithm are verified. Finally, we  C11 C12 C13 0 C15 0 
 
applied our proposed method to real data inversion and  C12 C 22 C 23 0 C 25 0 
achieved good results. C
13 C 23 C33 0 C35 0 
C   (5)
II. THEORY AND METHODS  0 0 0 C 44 0 C 46 
 
 C15 C 25 C35 0 C55 0 
A. Stiffness matrix of a group of inclined fractures with VTI
 0 0 0 C 46 0 C66  66
host rock 
Horizontal bedding is a layered structure of rocks formed by where matrix C describes a monoclinic medium whose the
primary sedimentation, and each layer is linear and parallel to vertical plane of symmetry coincides with the dip plane of the
each other. For the cases without fractures, shale reservoirs are fracture group. The approximate expressions of Cij are
usually regarded as VTI media [3], [32-33]. However, there are displayed in APPENDIX B.
often fractures in shale reservoirs [34-36]. For a group of
inclined rotationally invariant fractures with a VTI host rock, B. Linearized PP-wave reflection coefficient
the compliance matrix can be written as the compliance matrix According to Born scattering theory, Shaw and Sen [40]
of anisotropic background plus the excess compliance matrix of introduced the PP-wave reflection coefficient with the stiffness
fractures [28-29], [37]： matrix perturbation:
S  S b  S f  Cb1  S f  CVTI -1  M  S xf1 M T   (1) S r0 
RPP  (6)
where M   is a 6  6 rotation matrix, the specific definition 4  0 cos 2
is referred to Winterstein [38], T represents the matrix
x S r0   cos 2  C1111  C 22 22  C33 33  C 44 44 
transpose,  is the rotation angle, S f 1 is the excess compliance
C55 55  C 66 66  2C1212  2C1313  2C 23 23
of fractures, CVTI is the stiffness matrix of VTI background.
x (7)
The specific expressions of CVTI and S f 1 are shown in
Appendix A. Here we use the form of stiffness matrix

2
0196-2892 (c) 2021 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: Southwest Petroleum University. Downloaded on December 30,2021 at 15:55:55 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TGRS.2021.3138750, IEEE
Transactions on Geoscience and Remote Sensing
> IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING < 3

where  0 is the density of the background medium,  and [42], then the fracture weaknesses and velocities under the case
Cij (APPENDIX C) are perturbations of the density and of fluid saturation can be obtained.

stiffness coefficient matrix, respectively.  ij is defined in the

paper of Shaw and Sen [40]. By introducing the perturbation of
the stiffness matrix into equation (6), we obtain the linearized
reflection coefficient. In particular, the YPD approximate
equation is used as the isotropic part.
R pp  ,  ,    R iso
pp    R pp    R pp  ,  ,  
ani ani
(8)

E  
iso
RPP    C   
 D   
 E    (9)
E  

where, C     1 sec2   2g sin 2   ，

4  Fig. 1. Flow chart of rock physics modeling.

1  2g  3 2g  1 2 
1 2g  ，
D     sec 2   2g sin 2  In order to investigate the influence of the reflection
4 g 4g 3  3  4g  coefficient of the fracture inclination angle, a two-layer
 
medium model was established. The fluid in the fracture is
1 1 2 
E      sec   assumed to be a mixture of gas and water (water saturation is
2 4 . 25%). Given the VTI background stiffness coefficient, we
pp    F   T1  G   N1
R ani (10) study the variations of reflection coefficients with fracture
dipping angle (0o, 30o, 60o, 90o), incident angle (0o-40o) and
 
2
where, F    g sin2  , G     sec2  2g sin 2   1 azimuth (0o-180o) under different fracture densities (e=0, 0.02,
2 ， 0.04, 0.06, 0.08, 0.1) (Fig. 2). The PP-wave velocity, PS-wave
R ani
pp  ,  ,    A ,  ,   T  B ,  ,   N
2 2
(11) velocity and density of the upper medium are 3.41km/s,
A ,  ,    g sin 4  sin 4  - g sin 2  tan 2  
1.64km/s, and 2.6g/cm3 respectively. Suppose the background
stiffness coefficient (Gpa) of lower medium is expressed as
g sin 2
 sin 2   2 g sin 2  sin 4   g  g sin 2  cos 2  tan 2   [41]:
- g sin 2  cos 2   47.31 7.83 5.29 
 7.83 47.31 5.29 
(12)  
2

B  ,  ,    - g sin  sin   1 sin  tan  -
2 2

2 2 2  5.29 5.29 33.89
Cvti = 

 (14)
g sin   2 g  1 / 2g sin
2 2

  2 g sin 2  sin 2   1 / 2 tan 2  - 

17.15 

17.15
g sin   1 / 2
2 2



19.74 
(13)
where  is the incidence angle,  is the azimuthal angle, According to Fig. 2, we analyze the variations of amplitudes
 , E ,  are Poisson's ratio,Young's modulus and with fracture density, fracture dipping angle, incident angle and
   azimuthal angle, and draw the following conclusions:
 E 
(1)While the fracture density is 0, the reflection coefficient
density reflectivity, respectively. Poisson's ratio and Young's does not change with the azimuth. At this time, the medium
modulus can be used to calculate the brittleness of rocks and degenerates to the VTI medium.
guide the reservoir fracturing development.
(2)With the increase of fracture dipping angle, the reflection
C. Rock physics modeling analysis coefficient increases gradually (The upper medium remains
Rock physics research has built an effective bridge for unchanged).
fractured reservoir prediction using seismic data. Before (3)When there is a large dipping fracture in the medium, the
performing AVAz inversion, it is necessary to establish an anisotropy of the medium is stronger than that of the medium
anisotropic model to estimate the on-well weakness parameters. with a small dipping fracture, and the reflection coefficient
The rock physics modeling process is shown in Fig. 1. The changes more sharply with the azimuth.
background moduli can be obtained by mineral composition,
volume fraction and the porosity of matrix pores [44]. We add (4)With the increase of fracture density, the anisotropy of
inclined fractures to VTI background by linear slip model [39], fractured media increases as well as the reflection coefficient
and add fluid to fractures through Brown-Korringa formula difference with different fracture dipping angles. It
demonstrates that when the fracture density is high, replacing
inclined fracture with vertical fracture will bring some errors.

3
0196-2892 (c) 2021 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: Southwest Petroleum University. Downloaded on December 30,2021 at 15:55:55 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TGRS.2021.3138750, IEEE
Transactions on Geoscience and Remote Sensing
> IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING < 4

Fig. 2. The variations of reflection coefficients with fracture density (e = 0, 0.02, 0.04, 0.06, 0.08, 0.1), fracture dipping angle (0˚, 30˚, 60˚, 90˚), incident
angle (0˚-40˚) and azimuthal angle (0˚-360˚) under a given VTI background. (a) - (f) are fracture densities of 0, 0.02, 0.04, 0.06, 0.08 and 0.1, respectively.

D. AVOAz Inversion for fracture weaknesses, Poisson's ratio fracture weaknesses. Fig. 3 shows the influences of Young's
and Young's modulus modulus, Poisson's ratio, density and equivalent bedding
weaknesses on the PP-wave reflection coefficient. The
The linearized PP-wave reflection coefficient of the
influences of fracture weaknesses on the reflection coefficients
monoclinic medium can be characterized by Young's modulus,
under different fracture dipping angles are shown in Fig. 4.
Poisson's ratio, density, equivalent bedding weaknesses and

Fig. 3. The influence of inversion parameters (a: E / E , b:  /  , c:  /  , d:  N 1 and e:  T 1 ) on the PP-wave reflection coefficient. All
parameters range from - 0.1 to 0.1, and the arrow indicates the direction in which the parameter increases.

4
0196-2892 (c) 2021 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: Southwest Petroleum University. Downloaded on December 30,2021 at 15:55:55 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TGRS.2021.3138750, IEEE
Transactions on Geoscience and Remote Sensing
> IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING < 5

Fig. 4. The influences of the variations of fracture weaknesses on the reflection coefficient under different fracture dipping angles. (a) - (d) the influence
of normal weakness on reflection coefficient when fracture dipping angles are 0˚, 30˚, 60˚ and 90˚, respectively; (e) - (h) the influence of tangential
weakness on reflection coefficient when fracture dipping angles are 0˚, 30˚, 60˚ and 90˚ respectively.

Following Downton [47], the seismic azimuthal

In Fig. 3, we can observe that the contribution to the amplitude difference with N reflection interfaces and M
reflection coefficient of the parameters related to the VTI incident angles can be expressed as:
background does not vary with the azimuth. The reflection
 S 1, 1,    S 1, 2 ,     WA1,   WB1,   
coefficient increases with the increase of E / E ,  /  and      T2 
        
 /  , and decreases with the increase of  N 1 and  T 1 . S  M , 1,    S  M , 2 ,   WA M ,   WB M ,  

 N 2  2 N 1
MN1 MN2 N
Fig. 4 illustrates that when   0  , the influence of fracture (19)
weaknesses on reflection coefficient is consistent with that of
Equivalent bedding weaknesses. When   0  , the influence  w1 0  0
 1 
of fracture weaknesses on reflection coefficient varies not only  s1    w   
with the incident angle but also with the azimuth, and the   wk   0
contribution of normal weakness is greater than that of S     ,W    (20)
s N  0 wk  w1 
tangential weakness. The anisotropic parameters caused by  
fractures have less contribution to the reflection coefficient
  N 1
    
than other parameters, and the direct inversion of fracture  0  0 w k  N  N
weaknesses will cause instability. On the interfaces of
isotropic and anisotropic media, the weakness’s increment is where s1  s N are the elements of seismic data, w1  w k are
equal to weakness itself. Using equation (15), the influence of
the wavelet elements. Let m  T2 ,  N 2 T2 N 1 ,We can rewrite
the VTI background can be eliminated by the amplitude
difference [45-46] of two azimuths (   0  ,   90  ), which (19) as follows:
d  Gm (21)
can improve the inversion accuracy:
  
RPP  ,    R pp  ,   0 ,  - R pp  ,   90 ,   (15)
Based on Bayesian theory, assuming that the noise obeys
 A ,   T2  B ,   N 2
Gaussian distribution, combined with the smooth-model
regularization term and Cauchy constraint, the inversion

A ,    g tan 2  sin 2  cos 2  sin 2   cos 2   (16)
parameters can be obtained based on maximum a posteriori
(MAP) [47-49]. The objective function can be expressed as:
  
B  ,    tan 2  sin 2  g 2 sin 2  - 2 sin 2  - 2 - g   (17) F m   d  Gm T d  Gm   2 n2
2N

 ln1  mi2 /  m

2 


mod (22)
i 1
Based on formula (15) and convolution model, the
azimuthal amplitude difference of seismic data can associate 
T
  
with fracture weaknesses: mod  m  ηm  Pm R   ηm 1  Pm 1 R m  
1 1 1 m1
   1

  
S  ,    S  ,   0 ,  - S  ,   90 ,   (18) 
m 2  ηm 2  Pm 2 R
T
 
η  Pm R

(23)
 WA ,   T  WB ,   N  m 2   m 2 2 m2


2 2

5
0196-2892 (c) 2021 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: Southwest Petroleum University. Downloaded on December 30,2021 at 15:55:55 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TGRS.2021.3138750, IEEE
Transactions on Geoscience and Remote Sensing
> IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING < 6

where m i denotes the regularization coefficient of initial obtain the final inversion parameters. And we draw a
t1
flowchart to describe the inversion process (Fig. 5).
model, Pm i   d ，  m i  1 / 2 ln m i / mi 0  , mi 0 represents
t0

the initial model parameter. The derivation of equation (22) is

performed:
 T T T   T T T 
 G G  Q  m1 Pm Pm1  m2 Pm Pm2 m   G d  m1 Pm ηm1  m2 Pm ηm2 
 1 2   1 2 
(24)
 
1 1 1
Q  diag 
 2 2
,
2 2

2 2 
 (25)
 1  m1 /  m 1  m2 /  m 1  m2 N /  m 

where   2 n2 /  m2 ，  n2 represents the variances of seismic

noises，  m2 represents the variances of model parameters. Fig. 5. The flow chart of the proposed method.
Due to the introduction of Cauchy constraints, some
iterative algorithms [50-51] are no longer applicable. We use III. RESULTS
the iteratively reweighted least-squares optimization (IRLS)
algorithm [52] to solve equation (24). A. Synthetic data
After obtaining fracture weaknesses, we next invert Young's
modulus E , Poisson's ratio  , density  and equivalent Based on the rock physics modeling process (section II C),
bedding weaknesses  N1 and  T1 . the equivalent bedding weaknesses and fracture weaknesses
can be calculated by well logs. The reflection coefficient is
 S ' 1 ,  i ,     S 1 ,  i ,    calculated according to the given Young's modulus, Poisson's
 
      
  ratio, density, equivalent bedding weaknesses and fracture
 S  ,  ,  
'
 S  ,  ,   MN 1
 weaknesses, and the seismic record is synthesized with a 30
 M i  MN 1  M i Hz Ricker wavelet. Fig. 6 (a) shows the equivalent bedding
 WA 1 ,  i ,   WB 1 ,  i ,    weaknesses (  N1 and  T1 ) and fracture weaknesses (  N 2 and
   Tinv 
     inv 2
  GR ,  T2 ). Fig. 6 (b) shows young's modulus E , Poisson's ratio  ,
   
WA  M ,  i ,   WB  M ,  i ,    N 2  2 N 1
density  and fracture dipping angle  . In this study area,
MN  2 N
fractures have different dipping angles. In order to test the
(26)
stability of the proposed method, different Gaussian random
where, noises are added into synthetic traces. Under different signal to
 WC  1  WD  1  WE  1  WF  1  WG  
1 
noise ratios (No noise, SNR=5 and SNR=2), the azimuthal
  amplitude difference is obtained by subtracting two azimuthal
G       ，
seismic records (Fig. 7). Figures 8 (a), 9 (a) and 10 (a) show
WC M  WD M  WE M  WF  M  WG M  the inversion results of fracture weaknesses under different
MN 5 N
 T SNRs. Then, the residual parameters are inverted by using
 E    
R    T1  N1   . formula (28), and the elastic parameters Young's modulus,
 E    
 5 N 1 Poisson's ratio, density (Fig. 8 (b), 9 (b), 10 (b)) and equivalent
bedding weakness (Fig. 8 (c), 9 (c), 10 (C)) are obtained. The
where  Ninv2 and  Tinv 2
represent the inverted fracture theoretical data inversion results are affected by noises, but
weaknesses. The integration method in [17] can be used to have high stabilities.

Fig. 6. Well logs. (a) Equivalent bedding weaknesses and fracture weaknesses and; (b) Young's modulus, Poisson's ratio , density and dipping angle.

6
0196-2892 (c) 2021 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: Southwest Petroleum University. Downloaded on December 30,2021 at 15:55:55 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TGRS.2021.3138750, IEEE
Transactions on Geoscience and Remote Sensing
> IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING < 7

Fig. 7. Synthetic seismic amplitude differences of two azimuths.

(a) No noise, (b) SNR=5 and (C) SNR=2.

Fig. 8. The comparison between the true values and the inversion results (no noise). (a) The fracture weaknesses; (b) Young's modulus, Poisson's ratio and
density; (c) equivalent bedding weaknesses. The blue lines represents the initial parameters, the red lines are the inversion results, and the black lines are the
real values.

Fig. 9. The comparison between the true values and the inversion results (SNR=5). (a) The fracture weaknesses; (b) Young's modulus, Poisson's ratio and
density; (c) equivalent bedding weaknesses.The blue lines are the initial parameters, the red lines are the inversion results, and the black lines are the real values..

Fig. 10. The comparison between the true values and the inversion results (SNR=2). (a) The fracture weaknesses; (b) Young's modulus, Poisson's ratio, and
density; (c) equivalent bedding weaknesses. The blue lines are the initial parameters, the red lines are the inversion results, and the black lines are the real values.

B. Real data processing formation are well developed, with great potential for
three-dimensional exploration. The buried depth of the
To further prove the feasibility of the algorithm, real data reservoir is 2370-2520 m, with well-developed fractures and
have been processed. The study area is located in Sichuan good gas bearing property. The azimuthal angle  in the
Basin, where shale gas reservoirs of Wufeng-Longmaxi derived equation represents the difference between the

7
0196-2892 (c) 2021 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: Southwest Petroleum University. Downloaded on December 30,2021 at 15:55:55 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TGRS.2021.3138750, IEEE
Transactions on Geoscience and Remote Sensing
> IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING < 8

observed azimuth and the fracture strike. In the case of real parameters and logging data are also calculated (Table 1). It
data, the fracture strike should be determined before using the can be seen from Fig. 15 and Table 1, the inversion results are
observed seismic data for inversion. The formation consistent with logging data, which shows that the inversion
micro-imaging (FMI) data show that the horizontal bedding of results are stable and reliable.
formation is relatively developed, and there is a group of Finally, we extracted a slice along the target layer, which is
inclined fractures in the reservoir, the dipping angle is mainly shown in Fig. 16. From the slice, we can see that the fractures
concentrated in 70o. The input of the inversion algorithm and bedding are relatively developed around well.
including: two azimuth data of 0° and 90° to calculate azimuth
difference data at two incident angle (here we use 20° and 30°),
five stacked angle gathers (4°, 11°, 18°, 25°, and 32°), low
IV. DISCUSSIONS
frequency model of all parameters, and wavelets of different
angles. Before inversion, the seismic data should be processed This paper derives the reflection coefficient equation of
by the following steps: pre-stack denoising ( random noise, arbitrary fracture inclination in the VTI background, and uses
multiple waves, linear interference), surface consistency the azimuth difference data inversion algorithm to enhance the
processing, wavefront diffusion compensation, inverse Q stability of the inversion. The inversion of theoretical and
filtering, etc. Fig. 11 shows the azimuthal amplitude difference actual data shows that the algorithm has good applicability.
data at different incident angles (   20  and   30  ). Fig. 12 The characteristics of the research algorithm can be
shows the initial model used in the inversion process. summarized as follows:
The final inversion results of fracture weaknesses are shown (1)Compared with the reflection coefficient equations
in Fig. 13, and the location of target zone is marked by an derived by other scholars in the case of horizontal or vertical
ellipse. We observe that the fracture weaknesses have fractures in the VTI background, the reflection coefficient
relatively large values in the reservoir. The inversion results of equation derived in this study has wider applicability and can
Poisson's ratio, Young's modulus and equivalent bedding describe fractures at arbitrary inclination in the VTI
weaknesses are displayed in Fig. 14. Figures 14(e) and 14(d) background.
show that the stratum bedding is relatively developed in the (2)Through rock physics model and analysis, it is the first
reservoir location, which is a typical shale reservoir feature. time to obtain the influence of fracture inclination on the
Also, the target layer shows the characteristics of low reflection coefficient under the background of VTI. The
Poisson's ratio and low Young's modulus. analysis result shows that it is necessary to add the parameter
For facilitating the evaluation of the quality of the inversion of fracture inclination in the reflection coefficient equation.
results, Fig. 15 display the comparison between the borehole (3)The current research work does not discuss how to invert
side trace seismic inversion results with the logging data. And the fracture dip angle. When using the method in this paper, it
the correlation coefficients (CCs) between inversion is necessary to obtain the fracture dip angle of the reservoir
before inversion based on the FMI data.

Fig. 11. Azimuthal amplitude difference at different incident angles. (a)   20  ; (b)   30  . The black vertical line indicates the well
location and the ellipse indicates the reservoir location.

8
0196-2892 (c) 2021 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: Southwest Petroleum University. Downloaded on December 30,2021 at 15:55:55 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TGRS.2021.3138750, IEEE
Transactions on Geoscience and Remote Sensing
> IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING < 9

Fig. 12. Initial models of parameters to be inverted. (a) Young's modulus; (b) Density; (c) Poisson's ratio; (d) Normal equivalent bedding
weakness; (e) Tangential equivalent bedding weakness; (f) Normal fracture weakness; (g) tangential fracture weakness. The black vertical line
indicates the well location.

Fig. 13. Inversion sections of fracture weaknesses. (a) Normal fracture weakness; (b) tangential fracture weakness. The black curves are pseudo
well curves of fracture weaknesses, and the black ellipse marks the location of target zone.

9
0196-2892 (c) 2021 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: Southwest Petroleum University. Downloaded on December 30,2021 at 15:55:55 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TGRS.2021.3138750, IEEE
Transactions on Geoscience and Remote Sensing
> IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING < 10

Fig. 14. Inversion results of elastic parameters and equivalent bedding weaknesses. (a) Young's modulus; (b) Density; (c) Poisson's ratio; (d)
Normal equivalent bedding weakness; (e) Tangential equivalent bedding weakness. The black curves are well curves, and the black ellipse marks
the location of target zone.

Fig. 15. Comparison between borehole side trace inversion results and logging data. (a) Young's modulus; (b) Density; (c) Poisson's ratio; (d)
Normal equivalent bedding weakness; (e) Tangential equivalent bedding weakness; (f) Normal fracture weakness; (g) Tangential fracture
weakness. The blue curves are well curves, and the red curves are borehole side trace seismic inversion results.

Table 1 CCs between borehole side trace inversion results and logging data

Parameters E   N 1
T 1
N 2
T 2

CC 0.9755 0.9602 0.9285 0.9660 0.9621 0.9514 0.9615

10
0196-2892 (c) 2021 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: Southwest Petroleum University. Downloaded on December 30,2021 at 15:55:55 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TGRS.2021.3138750, IEEE
Transactions on Geoscience and Remote Sensing
> IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING < 11

Fig. 16. Slices extracted along the bottom interface of target layer. (a) Young's modulus; (b) Density; (c) Poisson's ratio; (d) Normal equivalent
bedding weakness; (e) Tangential equivalent bedding weakness; (f) Normal fracture weakness; (g) Tangential fracture weakness. Black circle
represents well position.

[9] V. Vavryčuk, and I. Pšenčík, “PP-wave reflection coefficients in weakly

anisotropic elastic media,” Geophysics, vol. 63, pp. 2129-2141, Nov. 1998.
V. CONCLUSIONS
[10] I. Pšenčík, and V. Vavryčuk, “Weak contrast PP wave displacement R/T
In this paper, the reflection coefficient equation of VTI coefficients in weakly anisotropic elastic media”, Pure appl. Geophys., vol.
medium reservoir with a group of inclined fractures is derived. 151, pp. 699-718, Mar. 1998.
Combined with Bayesian theory, the inversion accuracy of [11] M. Zillmer, D. Gajewski, and M. Kasthan, “Reflection coefficients for
shale reservoir parameters is effectively improved. In addition weak anisotropic media,” Geophys. J. Int., 129, pp. 389-398, Apr. 1997.
to the common reservoir parameters such as young's modulus, [12] M. Zillmer, D. Gajewski, and M. Kasthan, “Anisotropic reflection
Poisson's ratio and density, this method can also obtain the coefficients for a weak-contrast interface,” Geophys. J. Int., vol. 132, pp.
equivalent bedding weaknesses parameters representing the 159-166, 1998.
development degree beddings and the fracture weaknesses [13] P. Jilek, “Converted PS-wave reflection coefficients in weakly
parameters representing the fracture development degree, anisotropic media,” Pure appl Geophys., vol. 159, pp. 1527-1562, Jul. 2002.
which is of great significance to the effective development of [14] R. K. Shaw, and M. K. Sen, “Born integral, stationary phase and
shale reservoirs. The inversion results of theoretical and linearized reflection coefficients in weak anisotropic media,” Geophysical
Journal International, vol. 158, pp. 225-238, Jun. 2004.
practical data show that the research method has good
application effect. [15] A. Bakulin, and V. Grechka, I. Tsvankin, “Estimation of fracture
parameters from reflection seismic data-Part I: HTI model due to a single
fracture set,” Geophysics, vol. 65, no. 6, pp. 1788-1802, Nov. 2000a.
REFERENCES
[16] R. Bachrach, M. Sengupta, A. Salama, P. Miller, “Reconstruction of the
[1] B. Goodway, J. Varsek, and C. Abaco, “Anisotropic 3D amplitude layer anisotropic elastic parameter and high resolution fracture
variation with azimuth (AVAZ) methods to detect fracture prone zones in tight characterization from P-wave data: a case study using seismic inversion and
gas resource plays,” CSPG CSEG Convention, 2007. Bayesian rock physics parameter estimation,” Geophysical prospecting, vol.
[2] C. M. Sayers, “Seismic characterization of reservoirs containing multiple 57, pp. 253-262, Jan. 2009.
fracture sets,” Geophysical Prospecting, vol. 57, pp. 187-192, Feb. 2009. [17] X. Pan, and G. Zhang, “Fracture detection and fluid identification based
[3] A. Rüger, and I. Tsvankin, “Using AVO for fracture detection: Analytic on anisotropic Gassmann equation and linear-slip model,” Geophysics, vol. 84,
basis and practical solutions,” The Leading Edge, vol. 16, no. 10, pp. no. 1, pp. R99-R112, Jan. 2019.
1429-1434, Oct. 1997. [18] M. E. Far, “Seismic characterization of naturally fractured reservoirs,”
[4] A. Rüger, “P-wave reflection coefficients for transversely isotropic Ph.D. thesis, University of Houston, 2011.
models with vertical and horizontal axis of symmetry,” Geophysics, vol. 62, [19] M. E. Far, L. Thomsen, and C. M. Sayers, “Seismic characterization of
no. 3, pp. 713-722, Jun. 1997. reservoirs with asymmetric fractures,” Geophysics, vol. 78, no. 2, pp. N1-N10,
[5] A. Rüger, “Variation of P-wave reflectivity with offset and azimuth in Mar. 2013.
anisotropic media,” Geophysics, vol. 63, no. 3, pp. 935-947, May 1998. [20] V. Grechka, and I. Tsvankin, “Characterization of dipping fractures in a
[6] L. Thomsen, “Weak elastic anisotropy,” Geophysics, vol. 51, no. 10, transversely isotropic background,” Geophysical Prospecting, vol 52, pp. 1-10,
1954 -1966, Oct. 1986. Jan. 2004.

[7] J. Jech, and I. Pšenčík, “First-order perturbation method for anisotropic [21] E. Angerer, S. A. Horne, J. E. Gaiser, R. Walters, S. Bagala, and L. Vetri,
media,” Geophysical Journal of the Royal Astronomical Society, vol. 99, no. 2, “Characterization of dipping fractures using PS mode converted data,” 72nd
pp. 369-376, Nov. 1989. Annual International Meeting, SEG, Expanded Abstracts, pp. 1010-1013, Jan.
1999.
[8] A. Ruger, “Variation of P-wave reflectivity with offset and azimuth in
anisotropic media,” Geophysics, vol. 63, pp. 935-947, May 1998. [22] D. Nadri, J. Sarout, A. Bóna, and D. David, “Estimation of the anisotropy
parameters of transversely isotropic shales with a tilted symmetry axis,”
Geophysical Journal International, vol. 190, pp.1197-1203, Aug. 2012.

11
0196-2892 (c) 2021 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: Southwest Petroleum University. Downloaded on December 30,2021 at 15:55:55 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TGRS.2021.3138750, IEEE
Transactions on Geoscience and Remote Sensing
> IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING < 12

[23] J. Shragge, “Acoustic wave propagation in tilted transversely isotropic [44] R. Hill, “The elastic behavior of crystalline aggregate,” Proceedings of
media: Incorporating topography,” Geophysics, vol. 81, no. 5, pp. C265-C278, the Physical Society, vol. 65, pp. 349-354, Jan. 1952.
Sep. 2016.
[45] C. M. Sayers, J. E. Rickett, “Azimuthal variation in AVO response for
[24] H. Wang, and S. Peng, “Reflection coefficient of qP, qS and SH at a plane fractured gas sands,” Geophysical Prospecting, vol. 45, pp. 165-182, Oct.
boundary between viscoelastic TTI media,” Geophysical Journal 1997.
International, vol. 204, pp. 555-568, Nov. 2015.
[46] H. Chen, G. Zhang, Y. Ji, X. Yin, “Azimuthal seismic amplitude
[25] H. Chen T. Chen, and K. A. Innanen, “Estimating tilted fracture difference inversion for fracture weakness,” Pure and Applied Geophysics,
weaknesses from azimuthal differences in seismic amplitude data,” vol. 174, no. 1, pp. 279-291, Jan. 2017.
Geophysics, vol. 85, no. 3, pp. 1-58, May 2020.
[47] J. E. Downton, “Seismic parameter estimation from AVO inversion,”
[26] J. A. Hudson, “Overall properties of a cracked solid,” Mathematical Ph.D. thesis, University of Calgary, 2005.
Proceedings of the Cambridge Philosophical Society, vol. 88, pp. 371-384,
1980. [48] M. D. Sacchi, T. J. Ulrych, “High-resolution velocity gathers and offset
space reconstruction,” Geophysics, vol. 60, no. 4, pp. 1169-1177, Jul. 1995.
[27] J. A. Hudson, “Wave speeds and attenuation of elastic waves in material
containing cracks,” Geophysical Journal International, vol. 64, pp. 133-150, [49] W. M. Alemie, “Regularization of the AVO inverse problem by means of
Jan. 1981. a multivariate Cauchy probability distribution,” Master’s thesis, University of
Alberta, 2010.
[28] M. Schoenberg, “Elastic wave behavior across linear slip interfaces,”
Journal of the Acoustical Society of America, vol. 68, pp. 1516-1521, Nov. [50] S. J. Norton, “Iterative seismic inversion,” Geophysical Journal
1980. International, vol. 3, pp. 457-468, Jul. 1988.

[29] M. Schoenberg, “Reflection of elastic waves from periodically stratified [51] S. L. Pan, K. Yan, H. Q. Lan, J. Badal, Z. Y. Qin, “Adaptive step-size fast
media with interfacial slip,” Geophysical Prospecting, vol. 31, pp. 265-292, iterative shrinkage-thresholding algorithm and sparse-spike deconvolution,”
Apr. 1983. Computers and Geosciences, vol. 134, pp. 1-12, Jan. 2020.

[30] Z. Y. Zong, X. Y. Yin, F. Zhang, G. C. Wu, “Reflection coefficient [52] I. Daubechies, R. Devore, M. Fornasier, C. S. Güntürk, “Iteratively
equation and pre-stack seismic inversion with Young's modulus and Poisson reweighted least squares minimization for sparse recovery,” Communications
ratio,” Chinese J. Geophys., vol. 55, no. 11, pp. 3786-3794, Jan. 2012. on Pure and Applied Mathematics, vol. 63, no. 1, pp. 1-38, Jan. 2010.

[31] G. A. Waters, R. E. Lewis, and D. Bentley, “The effect of mechanical [53] G. E. Backus, “Long-wave elastic anisotropy produced by horizontal
properties anisotropy in the generation of hydraulic fractures in organic layering,” Geophysics, vol. 67, pp. 4427-4440, Oct. 1962.
shales,” SPE Annual Technical Conference and Exhibition, Society of [54] B. E. Hornby, L. M. Schwartz, and J. A. Hudson, “Anisotropic
Petroleum Engineers, 2011. effective-medium modeling of the elastic properties of shales,” Geophysics,
[32] A. Rüger, “Reflection Coefficients and Azimuthal AVO Analysis in vol. 59, pp. 1570-1583, Oct. 1994.
Anisotropic media,” Ph.D. thesis, Colorado School of Mines, 1996. APPENDIX A
[33] V. Grechka, I. Tsvankin, A. Bakulin, J. O. Hansen, C. Signer, “Joint
Based on the linear-slip model, the stiffness matrix of HTI
inversion of PIP and PS reflection data for VTI media: A North Sea case media has been written as [37], [39]:
study,” Geophysics, vol. 67, no. 5, pp. 1382-1395, Sep. 2002.
 M 1   N   1   N   1   N  0 0 0 
[34] A. Bakulin, V. Grechka, and I. Tsvankin, “Estimation of fracture  
parameters from reflection seismic data-Part II : Fractured models with   1   N  M 1   2 N   1   N  0 0 0 
   
orthorhombic symmetry,” Geophysics, vol. 65, no. 6, pp. 1818-1830, Nov.  
2000b. C HTI    1   N   1   N  M 1   2 N  0 0 0 
   
 0 0 0  0 0 
[35] A. Bakulin, V. Grechka, and I. Tsvankin, “Seismic inversion for the  0 0 0 0  1   T  0 
parameters of two orthogonal fracture sets in a VTI background medium,”  
Geophysics, vol. 67, no. 1, pp. 292-299, Jan. 2002.

 0 0 0 0 0  1   T 
66

[36] V. Grechka, I. Tsvankin, “Feasibility of seismic characterization of (A-1)

multiple fracture sets,” Geophysics, vol. 68, no. 4, pp. 1399-1407, Jul. 2003. where M    2  ,    / M .  ,  are the Lamé constants.
[37] M. Schoenberg, C. Sayers, “Seismic anisotropy of fractured rock,” The transformation between VTI and HTI media can be
Geophysics, vol. 60, no. 1, pp. 204-211, Jan. 1995. obtained by the matrix B  :

   1CHTI  B90 T 

[38] D. F. Winterstein, “Velocity anisotropy terminology for geophysicists,” 1
Geophysics, vol. 55, no. 8, pp. 1070-1088, Aug. 1990. CVTI  B 90 (A-2)
[39] M. Schoenberg, J. Douma, “Elastic wave propagation in media with
  
 M 1  2 g  1  N 
2  1  1  2 g  N   1   N 
parallel fractures and aligned cracks,” Geophysical Prospecting, vol. 36, pp. 0 0 0
   
571-590, Apr. 1988.   1  1  2 g   1  2 g  12    1   
N M  N N 0 0 0
   
[40] R. K. Shaw and M. K. Sen, “Use of AVOA data to estimate fluid CVTI    1   N   1   N  M 1   N  0 0 0
 
indicator in a vertically fractured medium,” Geophysics, vol. 71, no. 3, pp.  0 0 0  1   T  0 0
C15–C24, May 2006.  0 0 0 0  1   T  0 

 0 0 0 0 0  
[41] J. Guo, T. Han, L. ‐Y. Fu, D. Xu, X. Fang, “Effective Elastic Properties  66
ofRocks With Transversely Isotropic Background Permeated by Aligned (A-3)
Penny-Shaped Cracks,” Journal of Geophysical Research: Solid Earth,
vol.124, pp. 400-424, Jan. 2019.
In order to distinguish them from fracture weaknesses, here
we call  N ,  T in VTI medium as equivalent bedding
[42] R. Brown, J. Korringa, “On the dependence of the elastic properties of a
porous rock on the compressibility of the pore fluid,” Geophysics, vol. 40, no. weaknesses.
4, pp. 608-616, Aug. 1975. For a group of vertical rotationally invariant fractures, the
[43] A. W. Wood, A Textbook of sound, Mc Millan Co, New York, pp.360, excess compliance can be expressed as [15], [37]:
1955.

12
0196-2892 (c) 2021 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: Southwest Petroleum University. Downloaded on December 30,2021 at 15:55:55 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TGRS.2021.3138750, IEEE
Transactions on Geoscience and Remote Sensing
> IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING < 13

KN 0 0 0 0 0 
 
 0 0 0 0 0 0 
 0 0 0 0 0 0 
S xf1   (A-4)
 0 0 0 0 0 0 
 
 0 0 0 0 KT 0 
 0 0 0 0 0 K T  66

APPENDIX B
Stiffness Matrix Elements
The elements in equation (5) are listed below:
 
 2

C11  1  2g  12  N1 M - M  1 - 2g cos 2   N 2  4 g sin 2  cos 2  T2 

 (B-1)

 
C12  1  1  2g  N1  - M (1 - 2g ) 1 - 2 g cos 2   N 2   (B-2)
C13  1    - M 1  g
N1
2

sin 2 (2  )  2 g  N 2  g sin 2 (2  ) T2  (B-3)
C15   1  2 g sin   2
N2  cos 2  T2 sin 2  (B-4)

 2

C 22  1  2g  1  N1 M - M 2g - 1  N 2 2
(B-5)
 
C 23  1   N1  - M 2g - 1 2 g sin   1  N 2  2
 (B-6)
C 25   1  2 g  sin 2  N 2 (B-7)

 
  2

C 33  1   N1 M - M  2 g sin 2   1  N 2  4 g sin 2  cos 2  T2 

(B-8)


C 35   2 g sin 2   1 N2  cos 2  T2 sin 2  (B-9)
 
C 44  1   T1  -  cos 2  T2 (B-10)
1
C 46   sin 2  T2 (B-11)
2
C 55   
 1   T1  -  g sin 2 (2  ) N 2  cos 2 (2  ) T2  (B-12)
C 66   -  sin 2  T2 (B-13)
where  N1 an  T1 are equivalent bedding weaknesses,  N 2 and  T2 are fracture weaknesses.

APPENDIX C
Perturbation Matrix Elements
Under the assumptions of weak contrast and small weaknesses, the perturbation term of the stiffness matrix can be obtained by
neglecting the term proportional to  N i , M N i ,  Ti , M Ti ( i  1,2 ):

2

C11  M  M 2g  12  N1 - M  1 - 2g cos 2   N 2  4 g sin 2  cos 2   T2 
 
 (C-1)

C12  M  M 1  2g 2  N1 - M（1 - 2g）1 - 2 g cos 2   N 2   (C-2)


C13  M  2  M (1  2 g ) N1 - M 1  g 2 sin 2 (2  )  2 g  N 2  g sin 2 (2  ) T2   (C-3)

 
C15   1  2 g sin 2   N 2  cos 2   T2 sin 2   (C-4)

C 22  M  M 1  2g 2  N1 - M (2 g  1) 2  N 2 (C-5)

C 23  M  2 - M (1  2 g ) N1 - M 2g - 1 2 g sin 2   1  N 2   (C-6)

C 25   1  2 g sin 2   N 2 (C-7)

13
0196-2892 (c) 2021 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: Southwest Petroleum University. Downloaded on December 30,2021 at 15:55:55 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TGRS.2021.3138750, IEEE
Transactions on Geoscience and Remote Sensing
> IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING < 14

2
 
C33  M - M N1 - M  2 g sin 2   1  N 2  4 g sin 2  cos 2   T2 
 
(C-8)

 
C35   2 g sin 2   1  N 2  cos 2   T2 sin 2   (C-9)

C 44   -  T1 -  cos 2   T2 (C-10)

1
C 46   sin 2   T2 (C-11)
2

C55   -  T1 -  g sin 2 (2  ) N 2  cos 2 (2  ) T2  (C-12)

C66   -  sin 2   T2 (C-13)

APPENDIX D
Reflection coefficient equation under different assumptions
When the background is isotropic ( R ani    0 ), the reflection coefficient is degenerated to TTI medium, and is consistent with
the formula derived by Chen et al. [25].
R pp  ,  ,    R iso
pp    R pp  ,  ,  
ani
(D-1)

When   90 , the reflection coefficients of orthorhombic medium with a group of rotationally invariant fractures are obtained:

R ani  
2 2 2 2

pp  ,  ,   90  g sin  cos  1  tan  sin   T2   sec 2 
4
  2
2 g sin 2  cos 2   cos 2   1  N 2   (D-2)

which is consistent with the reflection coefficient equation derived by Pan et al. [39].
.

Ziyu Qin is currently pursuing the Ph.D. Bo Li is currently pursuing the Ph.D. degree
degree in earth exploration and information in earth exploration and information
technology at Chengdu University of technology at Chengdu University of
Technology, Chengdu, China. Technology, Chengdu, China.
Her research interest includes automatic His research mainly includes Q-value
first-arrival time picking, Seismic data extraction and Q-value inversion of seismic
Interpolation and AVAz (amplitude versus data.
incident and azimuthal angle) inversion.
Jingyi Chen is Decker Dawson Associate
Xiaotao Wen is professor of Geophysics in Professor of Geophysics at The University of
Chengdu University of technology. He is Tulsa. He is the director of Seismic
currently the vice president of the school of Anisotropy Group at The University of
Geophysics in Chengdu University of Tulsa.
technology. His research interests focus on both
His main research is reservoir geophysical conventional and unconventional reservoirs
comprehensive prediction and hydrocarbon including numerical simulation of seismic wave propagation,
detection. reverse time migration (RTM), full waveform inversion (FWI),
deep learning, Amplitude versus offset (AVO), Vertical
Dongyong Zhou received a Ph.D degree in seismic profile (VSP), microseismic, multiple-component
Earth exploration and information technology seismic data processing and reservoir characterization.
in China University of Petroleum (East China).
His main research direction is seismic data
inversion method.

14
0196-2892 (c) 2021 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: Southwest Petroleum University. Downloaded on December 30,2021 at 15:55:55 UTC from IEEE Xplore. Restrictions apply.