Journal of Applied Geophysics 116 (2015) 10–16

Contents lists available at ScienceDirect

Journal of Applied Geophysics

j ourna l homepage: www.e lsev ie r .com/ locate / j appgeo

Research on improving conventional marine seismic streamer dataimaging quality using OBS data from Northern South China Sea

Wang Xiangchun a,⁎, Han Youwei a, Liu Changchun a, Xia Changliang b

a Key Laboratory of Geo-detection, Ministry of Education, China University of Geosciences, Beijing, PR Chinab Overseas Business Department of Geophysical Research Institute, BGP Inc. of CNPC, Zhuozhou City, Hebei, PR China

⁎ Corresponding author.E-mail address: wangxc@cugb.edu.cn (X. Wang).

http://dx.doi.org/10.1016/j.jappgeo.2015.02.0240926-9851/© 2015 Elsevier B.V. All rights reserved.

a b s t r a c t

a r t i c l e i n f o

Article history:Received 18 November 2014Received in revised form 24 February 2015Accepted 24 February 2015Available online 26 February 2015

Keywords:OBSRayInvrStreamer dataMigration

It is very important to convert the seismic data from the time domain to the depth domain. Here we discuss theapproaches of inversemodeling of travel times for determination of the P-wave velocity (Vp) using OBS data, andalso using the inverted velocity to improve the conventional marine seismic streamer data imaging quality. Themigration section of the single channel seismic data is used to define themodel horizons and help to control theirgeometry. Wide angle hydrophone data of OBS are used to determine P-wave travel times. The picked traveltimes from various shots are inverted for P-wave interval velocities using RayInvr, which calculated theoreticaltravel times via ray tracing. Damped least squares optimization is performed to fine tune the fits between ob-served and calculated travel times. In the end, two conventional marine seismic streamer data's migration sec-tions are compared and the result shows that the section using the inverted velocity of OBS data is muchbetter than that using the velocity of conventional velocity analysis method to the streamer data.

© 2015 Elsevier B.V. All rights reserved.

1. Introduction

In the South China Sea continental margins, the presence of gas hy-drates has been reported (Chen et al., 2005; Han et al., 2008;Wang et al.,2012; Wu et al., 2005, 2007). Gas hydrate is a solid, ice-like crystallinesubstance formed from water molecules containing methane. Methanehydrates are stable under low temperature and high pressure, and aredetected on seismic data by identifying the characteristic bottom-simulating reflector (BSR), representing the base of the hydrate stabilityzone.

RayInvr (comprehensively described in Zelt and Smith, 1992)software has been used widely to invert the P- and S-wave velocitiesin the South China Sea (Wang et al., 2010; Zhao et al., 2010). Here wefirstly introduce the ray tracing principle and method, then use theRayInvr software to invert the P-wave velocity on five OBS hydro-phone data, including relocating the positions of the OBSs, definingmodel's horizons, picking the P-wave travel times, performing P-wave ray tracing inversion processes, and finally we got the invertedP-wave velocity/depth model. In the end, two conventional marineseismic streamer data's migration sections are compared and the re-sult shows that the section using the inverted velocity of OBS data ismuch better than that using the velocity of conventional velocityanalysis method to the streamer data. This illustrates that theinverted velocity is muchmore accurate than that of the convention-al velocity analysis method.

2. Data collection

The study area is located in the northern continental margin of theSouth China Sea (Fig. 1). Single-channel seismic (SCS) datawere collect-ed. In order to investigate the shallow sedimentary structures above theBSR, OBS data were also collected in this area in 2012 using the R/VHaiYang Liu Hao. By taking bandwidth output and bubble effect intoconsideration, a new type of GI gun point source was designed andthen applied to investigate the gas hydrate. The OBS instrument,MicrObs, is from the company Sercel. There are 39 parallel seismiclines shot normal to the continental margin with a distance of about25 m between two adjacent shots (inlet in Fig. 1). There are also 27lines shot parallel to the margin as tie lines (inlet of Fig. 1). For the 39NW–SE trending parallel lines, the spacing between two adjacent lineswas about 50 m, and five OBSs were deployed in water depths ofabout 1500 m along the central line (inlet of Fig. 1).

3. Streamer data processing

The followingworkflow (Fig. 2) is used to process the streamer data.Firstly the Segy format data is loaded and the geometry is set up. Sec-ondly spherical divergence compensation and predictive deconvolutionprocessing are used to the data. Thirdly velocity analysis is done and thevelocity is used to eliminate the multiple processing. In the end pre-stack time migration and Pre-stack depth migration processing aredone to the data. After the above processing, the pre-stack time migra-tion section (Fig. 3) and the Pre-stack depth migration section (Fig. 4)are achieved.

Fig. 1. The geometry of OBS and the shots: the red line is the main shot lines, the gray line is the ordinary shot lines, and the green line is the conventional streamer line.

11X. Wang et al. / Journal of Applied Geophysics 116 (2015) 10–16

4. Relocation OBS

In OBS seismic experiment carried out at sea, the real locations ofOBS on the seafloormay drift fromdesigned points (deployed locations)since OBSs are of free-fall type and usually affected by sea currents dur-ing their descent. So determining the OBS positions is a basic step forlater studies. Here the time-slice relocation method is presented andused to relocate the position of the ocean-bottom seismometer (OBS)on the sea bottom.

Fig. 2.Workflow of the conventional streamer data processing; the streamer data is proc-essed as this workflow: load data to processing software, set up geometry relationship ofshots and receivers on the software, do the spherical divergence compensation processingto improve the energy of the deep reflection events, do the predictive deconvolution pro-cessing to improve the vertical resolution, do the velocity analysis to achieve the velocityfield and do the demultiple to eliminate the multiples, do the pre-stack time migration toachieve the time migration section and do the pre-stack depth migration to achieve thedepth migration.

4.1. The principle of time-slice relocation method

When the velocity of seismic wave in the seawater is assumed to beconstant and the sea surface is not undulate, then the traveling time ofthe direct wave will be the same for the different shots which havethe same distances from the OBS. As seen in Fig. 5 (left), when theshots are on one circle, then the direct arriving wave's traveling timefor all shots is the same, and the distances from the shots to the projec-tion of the OBS on the sea surface are the same. According to this con-cept, we first determine the shots which have the same traveling timeof the direct wave, then the coordinates of all the shots are extracted,and it is shown that the shots are almost all on the same circle. Usingthe least squares method the center of the circle can be determined,and this can be seen as the horizontal position of the OBS.

One center of a circle can be achieved using one time slice, and whenwe use a number of time-slice data, a number of centers can be achieved.The final OBS horizontal position can be achieved by applying the statis-tics methods to the different circles' centers, as shown in Fig. 5 (right).

4.2. Relocation results

Table 1 is thedeployed and invertedOBSposition. Thefirst and secondcolumns are the x and y coordinates of the deployed position. The thirdand fourth columns are the x and y coordinates of the inverted position.

In order to verify the accuracy of relocation, linear move out correc-tion processing is used to the data. Fig. 6 (left) is the wave field of oneshot line after linearmove out correction processing using the deployedOBS position. Fig. 6 (right) is the wave field of one shot line after linearmove out correction processing using the time slice relocation method;when we calculate, ten time slice data is used, and the maximum dis-tance between each two inverted horizontal positions is 15 m. Becausethe gas hydrate lies in the layers near the ocean bottom, the accuracy ofthe OBS position is very important to the shallow layer's velocity inver-sion.When using the time slice relocation method, wemay see that thedirect arriving wave is almost horizontal so we can say that the reloca-tion result is reliable.

Fig. 3. Pre-stack time migration section, through processing of the streamer data as in Fig. 2's workflow, the time migration section is achieved as this one.

12 X. Wang et al. / Journal of Applied Geophysics 116 (2015) 10–16

5. P wave velocity inversion

5.1. Ray tracing principle and method

A linearized travel time inversion procedure, primarily developedfor modeling 2-D crustal refraction and wide angle data (applied in

Fig. 4. (Top) pre-stack depth migration section, through processing of the streamer data as indepth migration section using the inverted velocity of OBS data, compared to the migration se

Jose et al., 2008; Karastathis et al., 2001, 2002; Ogunsuyi et al., 2009;Parsiegla et al., 2009; Pim et al., 2008; Song and ten Brink, 2004), wasutilized in this study. Themodel is parameterized into a layered, irregu-lar arrangement of trapezoids to represent the velocity structure. Themodel parameters are boundary nodes (which, connected through lin-ear interpolation, define the structure of each layer boundary) and

Fig. 2's workflow, the depth migration section is achieved as this one. (Bottom) pre-stackction, the S/N ratio and the resolution are both improved.

Fig. 5. The principle of time slicemethod, first determine the shots which have the same traveling time of the direct wave; second the coordinates of all the shots are extracted, then usingthe least squares method the center of the circle can be determined. When a series of time slice is used, a series of centers can be achieved, and last the centers are averaged as the finalhorizontal position of the OBS.

Table 1The deployed and inverted OBS positions.

Station X (m) Y (m) X (m) Y (m)

1 696,872.20 2,436,189.70 696,850.90 2,436,188.852 697,081.90 2,435,860.88 697,047.34 2,435,861.083 697,291.66 2,435,532.06 697,293.50 2,435,558.344 697,501.40 2,435,203.27 697,525.86 2,435,256.775 697,711.17 2,434,874.42 697,752.51 2,434,851.06

13X. Wang et al. / Journal of Applied Geophysics 116 (2015) 10–16

upper and lower layer velocity points. The velocity field in each blockvaries linearly with depth (between the upper and lower velocities ina layer) as well as laterally across the velocity points along the upperand lower layer boundaries. Based on the work of Zelt and Smith(1992), rays are traced through velocity models in an iterative searchmode using zero-order asymptotic ray theory. Because the traveltimes between a source and a receiver depend on the model velocity,the inversion problem is a non-linear one. This is solved by linearizationusing a Taylor series expansion about a starting model and ignoringhigher terms, and then applying iterative analyses. Rays are tracedthrough the model based on zero-order asymptotic ray theory bysolving the ray tracing equations numerically (Cerveny et al., 1977)

Fig. 6. In the relocation results, linearmove out correction processing is used to the data: (left) isposition. The direct arriving wave is almost horizontal so we can say that the relocation result

using the Runge–Kuttamethod.When a ray intersects a layer boundary,which constitutes a velocity change, Snell's Law is applied.

5.2. Defining model horizons and P-wave travel time picking

Definition of model horizons is a key consideration when building arealistic velocity/depthmodel. Thus, it is expedient to gather all possibleinformation about the geology of the area and combine that with theavailable geophysical data to create a satisfactory model. Also, properseismic phase identification is important in order to avoid potentialgross errors in the final model (Zelt and Smith, 1992). Each significantpart of the subsurface must be presented in a realistic fashion by its in-terval velocity.

The single channel seismic profile data provide the basis for deter-mining model horizons. Firstly, according to the conventional streamerdata's depth migration section (Fig. 7) and the migration velocity field(Fig. 8 top), the initial velocity depth can be set up.

For this study, the aimed layer is the shallow layers that the gas hy-drate lies in, so we mainly inverted the shallow layers' velocity–depthmodel. The traveling time is two-way traveling time in the streamerdata's time migration section, and to the zero offset OBS data the

the result using the deployedOBS position and (right) is the result using the invertedOBSis reliable.

Fig. 7. Set up the initial depth model using the pre-stack depth migration section. The depth model is set up using the pre-stack depth migration section and also the 5 OBS are projectedonto the section.

14 X. Wang et al. / Journal of Applied Geophysics 116 (2015) 10–16

traveling time is one way traveling time. So when we compare the twodata, the traveling time of the OBS should be added a oneway travelingtime.

By combining the conventional streamer data's time migration sec-tion and the OBS data, we can pick up the responding reflection eventsof the OBS data (Fig. 9).

5.3. P-wave ray tracing

After setting up the initial velocity–depth model and picking thetravel times, ray tracing through the model was performed in a pro-gressive layer by layer fashion, as the velocities and depths of thelayers above directly influence the velocities and depths in question.Ray tracing is done by tracing through the model's trapezoids usingthe observed OBS travel time picks and the velocity/horizon depthcontrols specified in the input v. in file. The model was fine tunedusing RayInvr's damped least squares routine (dmplstsqr). The rou-tine uses travel time residuals and partial derivatives output by

Fig. 8. (Top) set up the initial velocitymodel pre-stackmigration velocityfield; the velocity is achthe depth domain. (Bottom) the inverted velocity model of OBS data; this is the final of the ve

RayInvr. These outputs are used by the routine to apply the dampedleast squaresmethodology of Aki and Richards (1980) and Lutter andNowack (1990) to solve a linearized inverse problem, and produce anew optimized model. This technique is employed in determiningthe changes in the model parameters, and the initial model is up-dated with the new values. Rays are then traced through the newmodel thereafter, and this procedure is repeated until a satisfactoryfit to the observed data is realized.

When we perform the inversion, the picked travel time error is animportant inversion parameter, here we give this error the same valueto all the horizons as 0.002 s for the minimum picked travel time errorwe can control is one time sample interval. Ray coverage through themodel is illustrated in Fig. 10. Ray coverage near the OBS is shown tobe very dense, constraining the velocity depth trade off. However, re-gions of the model outside the OBS array where few rays were tracedare not well constrained. It is worth noting that having the velocitynodes only at the beginning and end of the model produces an un-realistic velocity at the far lateral extremities of the model when the

ievedby the velocity analysis to the streamerdata and then convert it from time domain tolocity–depth model inverted form the OBS hydrone data.

Fig. 9. Combine the pre-stack time migration section and the OBS data to pick up the reflection events; the left part is the time migration section and the right is the OBS data.

15X. Wang et al. / Journal of Applied Geophysics 116 (2015) 10–16

layer is represented by a single trapezoid. Finally the inverted velocity–depth model is shown as Fig. 8 (bottom).

6. Improve streamer data imaging quality

The inverted velocitymodel is used to do the pre-stack depthmigra-tion again and the migration section is shown as Fig. 4 (bottom). Theother data processing is not changed. Compared to the migration sec-tion in Fig. 4 (top), the S/N ratio and the resolution are both improved.

Fig. 10. Ray coverage through the model, the velocity–de

This shows that the inverted velocity model is more accurate than thevelocity model derived from the streamer data.

7. Conclusion

The conventional marine seismic streamer data's migration sec-tion using the inverted velocity of the OBS data is much better thanthat using the velocity of the conventional velocity analysis methodto the streamer data. This illustrates that the inverted velocity is

pth inversion processing using the RayInvr software.

16 X. Wang et al. / Journal of Applied Geophysics 116 (2015) 10–16

much more accurate than that of the conventional velocity analysismethod.

Acknowledgments

This research is funded by the National Natural Science Foundation ofChina (No. 41304103), the Fundamental Research Funds for the CentralUniversities of China and the National 863 Plan Project (2013AA092501).

References

Aki, K., Richards, P.G., 1980. Quantitative Seismology: Theory and Methods vol. 1. WHFreeman and Co., San Francisco.

Cerveny, V., Molotkov, I.A., Psencik, I., 1977. Ray Method in Seismology. UniversityKarlova, Prague.

Chen, D., Huang, Y., Yuan, X., Cathles, L.M., 2005. Seep carbonates and preserved methaneoxidizing archaea and sulfate reducing bacteria fossils suggest recent gas venting onthe seafloor in the Northeastern South China Sea. Mar. Pet. Geol. 22, 613–621.

Han, X., Suess, E., Huang, Y., Wu, N., Bohrmann, G., Su, X., Eisenhauer, A., Rehder, G., Fang,Y., 2008. Jiulong methane reef: microbial mediation of seep carbonates in the SouthChina Sea. Mar. Geol. 249, 243–256.

Jose, T., Minshull, T., Westbrook, G., Nouze, H., Ker, S., Gailler, A., Exley, R., Berndt, C., 2008.A geophysical study of a pockmark in the Nyegga region, Norwegian Sea. Proceedingsof the 6th International Conference on Gas Hydrates.

Karastathis, V.K., Papamarinopoulos, S., Jones, R.E., 2001. 2-D velocity structure of the bur-ied ancient canal of Xerxes: an application of seismic methods in archaeology. J. Appl.Geophys. 47, 29–43.

Karastathis, V.K., Louis, J.F., Arvanitis, M., Karmis, P., 2002. Application of inverse ray-tracing seismic modelling techniques in the Ikaria airport site. J. Balk. Geophys. Soc.5, 107–114.

Lutter, W.J., Nowack, R.L., 1990. Inversion for crustal structure using reflections from thePASSCAL Ouachita experiment. J. Geophys. Res. 95, 4633–4646.

Ogunsuyi, O., Schmitt, D., Ahmad, J., Jason, F., 2009. Seismic travel time inversion to com-plement reflection profile in imaging a glacially buried valley. SEG Houston 2009 In-ternational Exposition and Annual Meeting, pp. 3675–3679.

Parsiegla, N., Stankiewicz, J., Gohl, K., Ryberg, T., Uenzelmann-Neben, G., 2009. SouthernAfrican continental margin: dynamic processes of a transform margin. Geochem.Geophys. Geosyst. 10 (3), 1–20.

Pim, J., Peirce, C., Watts, A.B., Grevemeyer, I., Krabbenhoeft, A., 2008. Crustal structure andorigin of the Cape Verde Rise. Earth Planet. Sci. Lett. 272, 422–428.

Song, J.L., ten Brink, U., 2004. RayGUI 2.0—a graphical user interface for interactive for-ward and inversion ray-tracing. US Geol Surv Open-File Rep, 2004-1426.

Wang, T., Yang, B., Deng, J., Lee, C., Liu, C., 2010. Seismic imaging of gas hydrates in thenorthernmost South China sea. Mar. Geophys. Res. 31, 59–76.

Wang, X., Xia, C., Liu, X., 2012. A case study: imaging OBS multiples of South China Sea.Mar. Geophys. Res. 33, 89–95.

Wu, S., Zhang, G., Huang, Y., Liang, J., Wong, H.K., 2005. Gas hydrate occurrence on thecontinental slope of the northern South China Sea. Mar. Pet. Geol. 22, 403–412.

Wu, S., Wang, X., Wong, H.K., Zhang, G., 2007. Low-amplitude BSRs and gas hydrate con-centration on the northern margin of the South China Sea. Mar. Geophys. Res. 28,127–138.

Zelt, C.A., Smith, R.B., 1992. Seismic travel time inversion for 2-D crustal velocity structure.Geophys. J. Int. 108, 16–34.

Zhao, M., Qiu, X., Xia, S., Xu, H., Wang, P., Wang, T.K., Lee, C.S., Xia, K., 2010. Seismic struc-ture in the northeastern South China Sea: S-wave velocity and Vp/Vs ratios derivedfrom three-component OBS data. Tectonophysics 480, 183–197 (Mar Geophys Res).