You are on page 1of 8

Journal of Geophysics and Engineering

Journal of Geophysics and Engineering (2020) 17, 89–96 doi:10.1093/jge/gxz093

Identifying and predicting multiples based on


spread of velocity spectrum
Xiaofeng Dai *, Lideng Gan and Hao Yang
Research Institute of Petroleum E&D, PetroChina, Beijing 100083, China

Downloaded from https://academic.oup.com/jge/article/17/1/89/5614456 by guest on 12 September 2021


*Corresponding author: Xiaofeng Dai. E-mail: dawnpeak0101@163.com

Received 20 April 2019, revised 28 August 2019


Accepted for publication 16 October 2019

Abstract
The genesis of internal multiples is complicated and identification is difficult as their velocities
are similar to the velocity of primaries, so their residual time is short. The existing conventional
methods for identifying multiples are mainly used on a single-common mid-point (CMP) or
single line, so the result is indistinct and inaccurate. The concept of velocity spread is proposed,
whereby multiples are identified by obtaining the lateral spread of energy clusters on the velocity
spectrum. The proposed method uses image segmentation to binarise the velocity spectrum,
obtain the spread of velocity, and identify and predict multiples on the plane via attribute spread
slicing. In a 3D seismic of Sichuan, the multiples predicted are in good agreement with the known
wells. The case shows that velocity spread analysis can quantitatively identify and predict
multiples interference, and is complementary to the existing multiples identification method. It
can be used to monitor the multiples suppression effect in seismic processing and analyse the
interference of multiples in interpretation. In addition, the method has good practicability for
evaluating the reliability of reservoir prediction results and reduces the risk of well location
deployment during exploration.
Keywords: seismic velocity spectrum, velocity spread, image segmentation, internal multiples

1. Introduction the reliability of seismic inversion and subsequent seismic


interpretation.
In seismic exploration, seismic waves propagate through sub-
Due to the complexity of internal multiples, it is difficult
surface formations. When there is more than one strong
to identify and attenuate multiples in seismic data process-
reflective interface, such as the surface of bedrock, uncon-
ing (Giovanni et al. 2015). Residual multiples have an even
formity or igneous rock, internal multiples usually occur
greater impact on the quality of seismic data and generate
(Berkhout & Verschuur 1997; David 2011).
great risks in oil and gas exploration and development. As
Under normal circumstances, internal multiples occur in
such, it is necessary to identify and evaluate the distribution
various exploration areas in China, especially in the Sichuan,
of internal multiples.
Tarim and Ordos Basins where there are many strong re-
At present, seismic processors and interpreters usually
flection interfaces, and multiples interference occurring be-
identify multiples or evaluate their suppression effect based
tween layers is more serious. Multiples interference makes it
on the velocity and periodic characteristics of the velocity
difficult to determine the velocity spectrum; it degrades im-
spectrum, stack profile and seismic pre-stack gathers (Song
age quality (Malcolm et al. 2007); it distorts the amplitudes,
et al. 2014). For example, in the pre-stack stage, there are low-
frequencies and phases of primaries and it decreases im-
velocity energy clusters in the velocity spectrum. Besides,
age resolution. It also affects seismic attribute extraction and
as the offset increases, pull-down events drop down on the

© The Author(s) 2019. Published by Oxford University Press on behalf of the Sinopec Geophysical Research Institute. This is an Open Access article distributed under the terms of 89
the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted reuse, distribution, and reproduction in any medium,
provided the original work is properly cited.
Journal of Geophysics and Engineering (2020) 17, 89–96 Dai et al.

gathers due to insufficient normal moveout (NMO). In the 2. Principle


post-stacking stage, multiples can be identified in the seis- The velocity spectrum usually refers to the stacked velocity
mic stack profile by their similar travel periods, and the struc- spectrum of the multiple coverage technique used in seis-
tures of their strong reflection interfaces and of the underly- mic exploration. According to the principle of NMO in seis-
ing strong reflection interface. mic wave reflection, a series of velocities are used to scan,
In recent years, data-driven multiple prediction and recog- and the NMO of the traces in the CMP gather are calcu-
nition methods have developed rapidly, such as Surface- lated and corrected to obtain a time–velocity–stack energy
related multiple elimination (SRME) (Dragoset et al. 2010), matrix. When the scanning velocity is correct, the NMO cor-
inverse-scattering (Weglein et al. 1997) and Marchenko rection of the primaries is complete. Meanwhile, the traces in
(Meles et al. 2015). However, these methods are still in their the CMP gather are stacked without moveout, so the stacking
early stages as a whole, and the reliability and practicability of energy is the strongest, and the velocity at this time is called
them need to be researched and tested further. So, it is diffi- the optimal stacking velocity. In contrast, the stacking energy
cult to apply to real seismic data effectively on a large scale. is weakened when the NMO velocity is not suitable.

Downloaded from https://academic.oup.com/jge/article/17/1/89/5614456 by guest on 12 September 2021


However, existing methods for identifying and predicting An effective application of the velocity spectrum is the
multiples have significant deficiencies. First, these methods analysis of multiples. The type and intensity of these multi-
mainly identify multiples based on the shape and appearance ples can be roughly judged by the low-velocity abnormal en-
of seismic reflections on seismic profiles, which are subjec- ergy clusters on the velocity spectrum. However, this method
tive, and the results are not convincing. Second, all the above can only be used to analyse a single seismic velocity spectrum
methods can only identify multiples at a single observation and its multiples on pre-stack gathers, so it cannot give the
point (a common mid-point (CMP) gather or velocity spec- overall degree of development and distribution of the multi-
trum) or an observation line (a seismic stack profile), so it is ples on the plane of a 3D seismic area.
impossible to determine the degree of occurrence and distri-
bution of multiples on the entire plane of 3D seismic area. In
fact, in some exploration areas, lateral transformation of the
2.1. Velocity spread concept
underground structure and stratigraphy happens very fast,
and the multiples on the plane are characterised by strong Normally, multiples and primaries have different propagation
heterogeneity. Obviously, the results of a certain CMP or a velocities. Therefore, when there is interference by multiples
single seismic line cannot represent the entire 3D seismic in the seismic data, the velocity spectrum usually indicates
area. In addition, compared to surface multiples, the velocity that the energy cluster is unfocused, and the velocity trend
of internal multiples is similar to the primary velocity and along the time axis is not unique.
the residual time is short. The multiples seriously inter- Figure 1 shows the velocity spectrum of a CMP in the Or-
fere with deep primaries, which makes identification more dos Basin, which shows that the multiples are not developed
difficult. in the shallow area above 2.0 s, the primary energy clusters
As such, the current method for identifying multiples has are well focused, the energy is concentrated and the range
low accuracy and cannot determine the development and of the stacking velocity corresponding to the energy clus-
distribution of multiples on the plane of an entire 3D seismic ters is narrow. For the time gate below 2.0 s, multiples grad-
area. It is very difficult to meet the requirements of internal ually develop. In the velocity spectrum, the energy gradually
multiples identification during seismic data processing and diverges and shows both a low-velocity trend (black dotted
interpretation. line), which indicates the reflection of multiples, and a high-
Considering these difficulties and the deficiencies of exist- velocity trend (red dashed line), which indicates the primary
ing methods, we propose a new method for identifying mul- effective reflection.
tiples based on the seismic velocity spectrum, which enables An analysis of the distribution characteristics of the en-
prediction of the development and scope of multiples on a ergy clusters indicates that the velocity spectrum shows a
plane based on the velocity spread along the layers. certain change rule due to the different degrees of multiples
Compared with the previous method for identifying mul- interference. When there are no multiples or only weak
tiples, our proposed method is highly efficient, is subject to multiples interference, the energy cluster is unique, the
little human interference, provides intuitive results and can concentration is good and the energy cluster is narrow in
identify and predict the degree of development and distri- the lateral direction. For example, at 2.0 s in figure 1, the
bution of multiples for an entire seismic area. The proposed energy cluster’s velocity ranges from 2230 to 2450 m s−1 ,
velocity spread identification method improves the ability for a velocity difference of 220 m s−1 . When the multiples
to identify multiples, enhances the method of multiples and primaries velocities are close in value, the two energy
identification and process monitoring, and demonstrates clusters are either superimposed or partially overlapped,
good practicability. the focus deteriorates and the width of the energy clusters

90
Journal of Geophysics and Engineering (2020) 17, 89–96 Dai et al.

is to use the threshold value for this determination. If it is


greater than a certain value, it is the effective energy, and vice
versa if it is invalid energy. However, due to the acquisition,
processing and techniques applied, the energy and signal-
to-noise ratios (SNRs) of the seismic data will change both
in the horizontal and vertical directions. The quality of the
velocity spectrum also varies greatly. Currently, if a single
threshold value method is applied to an entire 3D seismic
area, lots of errors are generated. As such, a more efficient
image binarisation method is needed for velocity spectrum
classification.

Downloaded from https://academic.oup.com/jge/article/17/1/89/5614456 by guest on 12 September 2021


3. Method for predicting multiples velocity spread
3.1. Velocity spectrum classification
In the image binarisation process, a binarised image is ob-
tained in which the overall and local features of the image
can be identified by selecting grayscale images with differ-
ent brightness levels depending on the threshold selection.
To obtain an ideal binary image, a closed, connected bound-
ary is generally used to define a non-overlapping region, and
Figure 1. Seismic velocity spectrum of a CMP in the Ordos Basin.
all pixels with grayscales greater than or equal to the selected
threshold are determined to be a specific object, the other
broadens. For example, the velocity range corresponding to pixels excluded from the object region are considered to be
the energy clusters at 2.3 s in figure 1 is 2600–2920 m s−1 , background or excluded object areas.
for a velocity difference of 320 m s−1 . When the velocity A 2D array of the velocity spectrum A(t, v) for any CMP
difference in multiple and primary increases, the two energy can be treated as grayscale image data. Using the chosen im-
clusters separate. For example, at 3.0 s in figure 1, the velocity age segmentation algorithm, the image segmentation thresh-
range of the two energy clusters is from 2000 to 3200 m s−1 , old is determined, and the velocity spectrum is binarised and
for a velocity difference increase of 1200 m s−1 . divided into background and target categories to realise class
When there is multiples interference, the stacking ve- segmentation of the velocity spectrum. The target category is
locity range corresponding to the energy clusters on the assigned a value of 1, which indicates that the seismic reflec-
velocity spectrum is widened. As the velocity difference tion forms an effective energy cluster along the velocity spec-
between the multiple and primary increases, the stacking trum; the background category is assigned a value of 0, which
velocity range widens. This feature can be used to identify indicates background values that include random noise and
multiples. For the convenience of explanation, in this paper, abnormal values.
we define the velocity interval corresponding to the effec- In this paper, we use the Otsu algorithm (Otsu 1979) as
tive energy cluster on the velocity spectrum as the velocity an example of the binarization process.
spread. The Otsu algorithm uses the interclass variance between
the target and background to determine the class. The seg-
mentation with the largest variance between classes has the
2.2. Basic principle smallest probability of misclassification. The interclass vari-
For a set of seismic traces at a CMP, the velocity spectrum can ance is calculated by setting each gray value as a thresh-
be represented by a two-dimensional arrayA(t, v), where t is old, and then the largest interclass variance use this as the
time, v is velocity and A is the velocity spectrum value (stack- threshold.
ing energy or correlation coefficient). The detailed calculation steps are as follows:
The start and end velocities corresponding to the effective
energy cluster are searched at the same time in the velocity (1) Sort the CMP velocity spectrum A(t, v) by value in
spectrum A(t, v). So, the velocity interval between the two is ascending, and record ci , i = 1, 2, ... , N as gray values;
the velocity spread at that moment. N being the total number of grid points, N = m × n,
The key to calculating velocity spread is determining how where m is the number of time sample points and n is
to identify the effective energy cluster. The simplest method the number of velocity sample points.

91
Journal of Geophysics and Engineering (2020) 17, 89–96 Dai et al.

(2) Calculate interclass variance 𝜎k of the CMP velocity Therefore, in seismic interpretation researches, a velocity
spectrum using equation (1): spread slice can be used to analyse the overall degree of de-
( N )2 velopment and distribution of multiples in the seismic data.
(N − k)k ∑ ci ∑k
ci The wider the range of the velocity spread slice, the greater
𝜎k = − . (1) the development of the multiples. The region with a large
N2 i=k+1
N − k i=1 k
spread value on the plane indicates the development of multi-
In equation (1), ci is the gray value, N is the total num- ples. Therefore, based on the velocity spread, it is possible to
ber of grid points and k is the order number. distinguish areas where seismic data are unreliable, provide a
When 𝜎k has its maximum value, the corresponding valuable reference for the reliability of interpretation results
gray value is the threshold value, which is Ck . and reduce the risk associated with the identification of well
deployment locations during exploration, thereby facilitating
(3) Using the gray value Ck as the threshold, divide the ve- oil and gas exploration and development.
locity spectrum into two categories, background 0 and In addition, in seismic data processing, a comparison

Downloaded from https://academic.oup.com/jge/article/17/1/89/5614456 by guest on 12 September 2021


target 1, using equation (2): of the spread slices of pre- and post-multiples suppression
{ reflects the degree of multiples suppression achieved. It also
1 Ai,j ≥ Ck assists in the analysis of the processing of multiples suppres-
Ti,j = . (2)
0 Ai,j < Ck sion, and guides the optimisation of the processing parame-
ters to effectively suppress multiples in seismic data, reduce
In equation (2), Ti,j is the velocity spectrum category, uncertainty, realise parameter optimisation, enable quantita-
Ai,j is the velocity spectrum value, Ck is the thresh- tive monitoring and improve the SNR and imaging effects.
old value and i, j indicate the number of time sampling
points and velocity sampling points, respectively.
3.4. Model test and analysis

3.2. Calculate velocity spread In this paper, model calculation is performed by the seismic
forward modelling of a known well, which is then used to
For the velocity spectrum class Ti,j gathers, the velocity analyse the influence of multiples on the velocity spectrum
spread is calculated point by point in chronological order. and the characteristics of velocity spread.
For each time i, a one-dimensional array is composed of The frequency-domain reflectivity is a useful forward
Ti,1 , Ti,2 , Ti,3 , ... , Ti,n . modelling method (Gan et al. 2018). In addition to the pri-
The following calculation method is used to obtain the ve- maries, multiples and converted waves can be considered si-
locity spread at each point in time in the array: for the array multaneously. Using the logs of a well in the Sichuan Basin,
Ti,1 , Ti,2 , Ti,3 , ..., Ti,n , the positions of the initial and final cat- we simulated the pre-stack primary and full-wave gathers
egory 1 values are recorded, and the velocity difference be- (primary + multiple) by the frequency-domain reflectivity
tween these positions is the velocity spread at that time. method. Figure 2 shows the forward modelling gathers of the
well. Next, the velocity spectra were calculated separately and
the velocity spread computed. We note that primary veloc-
3.3. Velocity spread identification of multiples
ity spectra only exist above 2.45 s, since the logging ends at
For the velocity spread volume, the velocity spread attribute 2.45 s.
is obtained for a certain time window. After the spread slice A comparison of the primary and full-wave gathers shows
is obtained, the multiples on the plane are identified and a that the multiples are mainly developed below 2.1 s, and the
prediction is made. energy of the multiples within the local range of 2.3–2.45 s is
The primary reflection is mainly affected by stratum fac- strong (shown in the rectangle in figure 2). In the time win-
tors. Therefore, in the same geological layer, the seismic data dow of multiples development, a low-velocity anomalous en-
have similar reflection characteristics, the velocity spectra ergy cluster is evident on the velocity spectrum of the full-
have good similarity and the velocity spreads are similar. If wave gathers. Compared with the primary velocity spectrum,
there are multiples originating from the shallow interface, the focus of the energy cluster is poor and the stacking ve-
in the seismic velocity spectrum, energy clusters are also locity range is widened, which are typical features of a multi-
formed in a low-velocity region that is lower than the primary ple velocity spectrum. Correspondingly, the full-wave veloc-
velocity, so that the lateral distance increases between en- ity spread value is significantly larger than that of the primary
ergy clusters in the velocity spectrum and the velocity spread during the time window of multiples development. Figure 2c
increases. Moreover, the greater the variation of the spread, shows the residual curve of the primary and full-wave veloc-
the greater the difference between the multiples and the pri- ity spread. This curve directly reflects the stacking velocity
maries velocity. of the multiples by subtracting the velocity of the primary.

92
Journal of Geophysics and Engineering (2020) 17, 89–96 Dai et al.

Downloaded from https://academic.oup.com/jge/article/17/1/89/5614456 by guest on 12 September 2021


Figure 2. Comparison of primaries and multiples forward modelling of a well.

Therefore, in theory, on the residual spread curve, there is and primaries make it difficult to identify and suppress the
no multiples region and the residual is zero, whereas in the multiples and the SNR of the seismic data is low (Yang et al.
region where multiples develop the residual is positive. The 2017; Dai et al. 2018; Dai & Xu 2018).
results of this model test calculation prove that the velocity
spread value can better indicate the region of multiples de-
velopment and enable quantitative prediction. 4.2. Workflow
First, the CGG seismic processing system is used to obtain
4. Application example the velocity spectrum gathers of the 3D seismic area in an
SGY format and the grid of the velocity spectrum analysis is
4.1. Geological background
400 × 400 m (the actual trace interval is 20 × 20 m).
The three-dimensional (3D) seismic area of this study is lo- Figure 3 shows the velocity spectrum of a CMP in the
cated in the central Sichuan Basin. High-yield and enriched study area. The color in the figure indicates the correlation
carbonate karst gas reservoirs have been discovered recently coefficient value, which ranges from 0 to 1; the abscissa is the
in the Sinian Dn Fm. in this area. The top of the Dn Fm. is at relative velocity percentage, which ranges from 80–120% and
a depth of 5000 m, and the thickness of the Dn Fm. is about the relative velocity change interval is 1%. That is, the lateral
800 m. velocity grid number m = 41; the ordinate is time, the time
This area is dominated by marine carbonate rock plat- interval is 4 ms and the time period ranges from 1.2 to 3 s. The
form deposits from the Sinian to Middle Triassic. During this number of longitudinal time grids is n = 450. Therefore, the
age, four sets of low-velocity shale and high-velocity carbon- velocity spectrum at this CMP includes 41 × 450 = 18 450
ate rock were developed and strong reflective interfaces were grid point values.
formed. Studies have shown that these strong reflection inter- Next, we classify the velocity spectrum by the above
faces cause strong multiples interference in the vicinity of the method to obtain a velocity spectrum class gather. Figure 4
Dn Fm. and the Lower Cambrian Qzs Fm. Due to the strong shows the results of the velocity spectrum classification at
multiples interference, the similar velocities of the multiples the CMP. In the figure, the white grid points represent

93
Journal of Geophysics and Engineering (2020) 17, 89–96 Dai et al.

Downloaded from https://academic.oup.com/jge/article/17/1/89/5614456 by guest on 12 September 2021


Figure 3. Seismic velocity spectrum of a CMP. Figure 5. Velocity spread of a CMP.

Figure 4. Velocity spectrum classification of a CMP.


Figure 6. Dn Fm.’s spread-attribute map of 3D area.

category 1 and the effective energy clusters on the velocity


spectrum; the black grid points represent category 0, which with the current understanding of the strong multiples inter-
indicate background noise. ference in the Cambrian and Sinian Dn Fm. in the study area.
Next, using the velocity spectrum class gathers, we cal- Finally, we calculated the average spread value along the
culate the velocity spread of all the CMPs, and obtain the Dn Fm., performed median filtering to suppress any abnor-
velocity spread volume of the 3D area. Figure 5 shows the mal values and obtained the velocity spread attribute slice.
velocity spread for the same CMP. As shown in figure 5, as
time increases, the velocity spread gradually increases. The
4.3. Effectiveness
maximum spread occurs between 2.4 and 2.8 s, which means
that there are different energy clusters and multiples develop- The spread-attribute map of the Dn Fm. in the study
ments in this time range at this CMP. This result is consistent area (figure 6) shows that the velocity spread has strong

94
Journal of Geophysics and Engineering (2020) 17, 89–96 Dai et al.

Figure 8. Cross-plot of correlation coefficient of well–seismic data and ve-

Downloaded from https://academic.oup.com/jge/article/17/1/89/5614456 by guest on 12 September 2021


locity spread.

and c, at 2.1 s, the consistency of the event characteristics of


the synthetic seismogram and the actual seismic trace deteri-
orates gradually. This occurs especially in the vicinity of the
Dn Fm. Except effective comparison of the Cambrian bottom
Figure 7. Well W6 synthetic seismogram calibration (∈1 q) peak, the time, energy and phase of other events is dif-
Note: ∈1 q-Cambrian Qzs Fm. and Z2 d-Sinian Dn Fm. ficult to correlate (the position indicated by the blue arrow in
figure 7). The calibration below 2.2 s is very poor, with a cor-
heterogeneity on the plane, which indicates an uneven spatial relation coefficient of only 0.14. The well calibration results
distribution. Near the gas enrichment area of the Dn Fm., the indicate that the seismic data contain strong multiples inter-
strong multiples interference is mainly located around wells ference, which is consistent with the highest velocity spread
W2 and W6, where the velocity spread is higher than 8. at the well location.
To verify the reliability of the multiple prediction, we cal- We verified the result of the multiples prediction based
ibrated the seismic with wells and compared the velocity on the velocity spread by well–seismic data calibration of all
spread attributes and correlation coefficients of the seismic wells in the 3D area region. The synthetic seismograms of
well. If there was no multiples interference near the well loca- all 10 wells are not highly correlated with the actual seismic
tion, the seismic data would have a high SNR, the synthetic traces, having an average correlation coefficient of only 0.56.
seismograms of the well and the seismic data should be in The variation between wells is great, with the correlation co-
good agreement and have a high correlation coefficient. Con- efficient ranging from 0.14 to 0.75 (figure 8). This demon-
versely, if multiples develop, the synthetic seismograms of the strates that the multiples interference in the Dn Fm. is seri-
well and the seismic data would be poorly matched and have ous and the lateral distribution is uneven, which is consistent
a low correlation coefficient. with the overall plane rules of the multiples predicted by the
As an example for analysis, we take well W6 near where a velocity spread. The correlation of well–seismic and the ve-
strong multiples interference zone has been predicted based locity spread have a good negative linear correlation, with a
on the velocity spread. This well is in the south part of the coincidence rate as high as 87% (figure 8). The greater the
study area at a depth of 5455.0 m. Based on a caliper mea- velocity spread value, the more intense the multiples inter-
surement, the log quality of the deep portion of the well is ference and the poorer the match between the seismic data
good and the sonic and density curves of the well are reliable. and the synthesis. This result provides sufficient proof that
Figure 7 shows a calibration diagram of well W6, where the velocity spread can effectively predict the level of multi-
figure 7a shows the impedance curve of the W6 well calcu- ples interference.
lated from the sonic and density logs; figure 7b shows the
11 seismic traces near W6, with the blue dotted line showing
5. Conclusion
the drilling trajectory; figure 7c is well W6’s synthetic seismo-
gram obtained by the convolution of seismic wavelet and the The concept of velocity spread is proposed in this paper. The
reflection coefficients calculated from the impedance curve; degree of multiples occurring on the plane can be predicted
and figure 7d shows the correlation coefficients obtained by by calculating the lateral velocity spread of the energy clus-
the cross-correlation of the 11 seismic traces in the figure 7b ters on the seismic velocity spectrum. The approach over-
and c synthetic seismograms of the well. Comparing figure 7b comes the deficiencies of single-point, single-line multiples

95
Journal of Geophysics and Engineering (2020) 17, 89–96 Dai et al.

identification, which is complementary to the multiple iden- Brookes, D., 2011. Case studies in 3D interbed multiple attenuation, The
tification and prediction method. Leading Edge, 30, 914–918.
The work flow of velocity spread analysis technology in- Dai, X.F., Liu, W.D. & Gan, L.D., 2018. The application of radon trans-
form to suppress interbed multiples in Gaoshiti-Moxi area, Acta Petrolei
cludes: first, binarisation processing the velocity spectrum Sinica, 39, 1028–1036.
by using the image segmentation technique; second, obtain- Dai, X.F. & Xu, Y.P., 2018. The suppression of Meso-NeoProterozonic
ing the volume of the velocity spread and last, predicting the deep multiples in central Sichuan, Chinese Journal of Engineering Geo-
quantitative distribution of multiples on the plane by extract- physics, 15, 189–194.
ing spread-attribute slice. Dragoset, B., Verschuur, E. & Moore, I., 2010. A perspective on 3D surface-
related multiple elimination, Geophysics, 75, 245–261.
With the application of real seismic data, velocity spread
Gan, L.D., Xiao, F.S. & Dai, X.F., 2018. Breakthrough and signif-
analysis technology is demonstrated with higher accuracy icance of technology on internal multiple recognition and sup-
and practicability. It can be used to monitor multiples sup- pression: a case study of Sinian Dengying Formation in Central
pression to eliminate multiples better at the seismic data pro- Sichuan Basin, SW China, Petroleum Exploration and Development, 45,
cessing stage, and it can be used to analyse multiples interfer- 960–971.

Downloaded from https://academic.oup.com/jge/article/17/1/89/5614456 by guest on 12 September 2021


ence of seismic data at the seismic interpretation stage, which Giovanni, A.M., Katrin, L. & Matteo, R., 2015. Internal multiple prediction
and removal using Marchenko autofocusing and seismic interferometry,
helps to evaluate the reliability of reservoir prediction results Geophysics, 80, A7–A11.
and reduce the risks associated with well deployment. Malcolm, A.E., De Hoop, M.V. & Calandra, H., 2007. Identification of im-
age artifacts from internal multiples, Geophysics, 72, S123–S132.
Meles, G.A, Löer, K & Ravasi, M., 2015. Internal multiple prediction
Acknowledgements
and removal using Marchenko autofocusing and seismic interferometry,
This work was supported by the National Key R&D project of Geophysics, 80, A7–A11.
China (2018YFC0603706) and CNPC project (2017D-3503). We Otsu, N., 1979. A threshold selection method from gray-level histograms,
would like to thank the E&D Research Institute of PetroChina IEEE Transactions on Systems, Man and Cybernetics, SMC-9, 62–
Southwest Oil & Gas field for providing the seismic data. 66.
Song, J.W., Verschuur, D.J. & Chen, X.H., 2014. Research status and
progress in multiple elimination, Progress in Geophysics, 29, 240–
Conflict of interest statement. None declared. 247.
Weglein, A.B., Gasparotto, F.A. & Carvalho, P.M., 1997. An inverse-
scattering series method for attenuating multiples in seismic reflection
References data, Geophysics, 62, 1975–1989.
Berkhout, A.J. & Verschuur, D.J., 1997. Estimation of multiple scattering Yang, H., Dai, X.F. & Gan, L.D., 2017. Reflectivity forward modeling and
by iterative inversion, Part 1. Theoretical considerations, Geophysics, 62, application of multiples recognition - a case study of Central Sichuan
1586–1595. Basin, CPS 2017 Geophysical Technology forum, 310–313.

96

You might also like