Marine broadband case study offshore China

first break volume 29, September 2011 technical article Marine broadband case study offshore China Tim Bunting, 1* Bee Jik Lim, 2 Chui Huah Lim, 3 Ed Kragh, 4 Gao Rongtao, 1 Shao Kun Yang, 5 Zhen Bo Zhang, 5 Yu Hong Xie 6 and Lie Li 6 Abstract Dual tow-depth acquisition configurations are used to mitigate the band-limiting effect of the sea-surface ghost in marine streamer surveys. Here we report on the analysis of a 2D broadband marine seismic project acquired offshore China. The deployment configuration included over/under sources and three streamers towed at different depths (5, 17, and 23 m). This configuration allowed not only for analysis of dual tow depth as a method for increasing bandwidth, but also for evaluation of two distinct dual tow-depth combinations, over/under and sparse-under. The project provided the first opportunity to make this comparison. Evaluation of reflection images, filter panels, amplitude spectra, and signal/noise separations demonstrates that both techniques are effective at compensating for the filtering effect of the ghost response, but also suggest that the improved results from each dual tow-depth technique are very comparable. Given the operational advantages of the sparse-under technique, due to the reduced streamer requirements, we conclude that the sparse-under technique is an attractive option for 3D deployments. Introduction The effect of the sea-surface ghost on marine seismic datasets is well understood. The ghost (sea-surface reflection) constructively and destructively interferes with the primary wavefield and acts to band-limit the recorded seismic measurement. This band-limiting effect has been a challenge since the inception of the marine seismic method. As a consequence, surveys are designed to optimize imaging of specific targets. For example, to optimize imaging of shallow targets, it is general practice to tow sources and streamers shallow. This practice results in good shallow resolution because the high frequencies are preserved, but at the expense of aggressive attenuation of the low frequencies. Ideally, we want data that are rich in both high and low frequencies to image shallow and deep targets optimally, and for seismic attribute analysis. Figure 1 shows the effect of the receiver ghost on the amplitude response for 5 m and 23 m tow depths in the frequency domain. The 5 m tow depth response is good at high frequencies, but has heavy attenuation at low frequencies, whereas the deep-tow response is good at low frequencies but has notches within the general seismic bandwidth. In recent years, two specific tow-depth combination solutions have been deployed, over/under and sparse-under, to solve for the band-limiting effects of the sea-surface ghost. Both configurations use streamers that are towed at different depths below the sea surface. The over/under technique (e.g., Hill et al., 2006) has been around for at least 50 years, but has gained wider acceptance over the last five years, since streamer steering technologies have been developed. Streamer steering minimizes the lateral separation of the two streamers, which can result from different water currents at the different tow depths, and is important because over/under wavefield separation techniques rely on the two streamers measuring the upgoing and downgoing wavefields at the same horizontal position. With the over/under technique, both streamers are towed deep, generally between 15 m and 30 m. The specific depths are chosen to ensure that there are no overlapping ghost notches in the bandwidth of interest. The measurements from the two streamers are combined post-acquisition using, for example, a de-phase and weighted-sum wavefield separation technique (Posthumus, 1993). Sparse-under is a more recent development. Initially described by Kragh et al. (2009), the technique complements a measurement from streamers towed at conventional depths (5 8 m) with measurements from a smaller number of deep streamers (15 30 m). Energy from the 1 WesternGeco, Brazil Research & Geoengineering Centre, Rua Paulo Emido Barbosa 485, Quadra 7B-Parque Technológico, Cidade Universitária URFJ, Isla da Fundao, Rio de Janeiro, Brazil. 2 WesternGeco, Level 5, 256 St Georges Terrace, Perth, Australia 6000. 3 WesternGeco, Rohas Perkasa 7 th Floor, No 8 Jalan Perak, Kuala Lumpur, 50450, Malaysia. 4 Schlumberger Cambridge Research, High Cross, Madingley Road, Cambridge CB3 0EL, UK. 5 CNOOC, Tower B, Offshore Petroleum Building, #1, Industry Road 2, Shekou, Shenzhen, 518067, China. 6 CNOOC, PO Box 22, Potou, Zhanjiang, Guangdong, 524057, China. * Corresponding author, E-mail: tbunting@slb.com 2011 EAGE www.firstbreak.org 67

technical article first break volume 29, September 2011 Figure 1 Shallow-tow (left) and deep-tow (right) streamer amplitude responses, showing the band-limiting effect of the sea-surface reflection. Figure 2 Block diagram of the optimal de-ghosting workflow in 2008 (left), and the low-frequency boosting workflow (right). The low-frequency boosting workflow incorporates the optimum de-ghosting combination within a more complex workflow. deep streamers provides a boost to the low frequencies of the shallow streamer measurement, which are heavily attenuated by the shallow-tow response. As the use of the deep streamers is limited to low frequencies, the deep streamers can be sampled more sparsely when deployed in a 3D configuration. This sparse sampling of the deep streamer plane results in acquisition efficiencies compared to 3D over/under configurations because of the reduction in streamers required to cover an equivalent subsurface swath. The combination method employed in each technique is necessarily different. Here we combine the over/under data using an extension of the de-phase and weighted-sum wavefield separation technique (Posthumus, 1993), referred to as optimal de-ghosting, to generate a ghost-free upgoing wavefield (Özdemir et al., 2008). This technique additionally considers the different noise responses at the different tow depths when computing the summation weights to optimize the signal-to-noise ratio in the combined response. The basic workflow (Figure 2) is:- 1) De-phase: convert the receiver component of the total response to zero phase by correlating with the individual depth ghost responses. 2) Sum: combine the two wavelets after correcting to a common vertical datum. 3) Spectral shaping: deterministically shape to the true upgoing receiver wavefield, free of receiver ghosts, by applying the inverse of the summed ghost response. Sparse-under summation also uses a similar de-phase and sum approach (Figure 2), but in a more complex workflow, essentially applying the combination only to the low frequencies while maintaining a constant wavelet phase over the whole bandwidth. This is achieved as follows. The shallow and deep datasets are de-phased and summed but 68 www.firstbreak.org 2011 EAGE

first break volume 29, September 2011 technical article the summation output is low-pass filtered. The shallow streamer dataset is de-phased in the same way to avoid any frequency blending artifacts prior to applying a high-pass filter. For this project, we ran the complete full-bandwidth 5 m dataset through the de-phase and sum workflow, i.e., both inputs use the same 5 m dataset. This converts the receiver component of the response to zero phase and also inverts the ghost response by the inherent spectral shaping. There is only so much that can be achieved with this single streamer de-ghosting approach because it is not possible to shape through the notch. However, since the high-frequency notch is outside the bandwidth of interest, a targeted bandwidth approach is perfectly acceptable. In August 2009, two 2D surveys were acquired offshore China. For both surveys, three streamers were deployed at different depths. The choice of streamer depths allowed the two different combination solutions to be evaluated against each other and against single depth measurements. The results demonstrate that both solutions have significant uplift in bandwidth and general signal-to-noise levels, but also demonstrate that, while the two solutions are very comparable in terms of geophysical quality, there are operational differences that may direct the user to choose one method or the other. Project background The two surveys were acquired for China National Offshore Oil Corporation (CNOOC), Shen Zhen (CCLS), and CNOOC Zhan Jiang (CCLZ) in a deepwater frontier area of the South China Sea, east of Hainan Island and south of Hong Kong (Figure 3). In total, five lines were acquired totalling over 1000 km of 2D subsurface profile. The two surveys were approximately 400 km apart in very different geological settings. For the CCLS survey, to the north-east, the shallow overburden was very complex. There was a significant amount of out-of-plane energy that would have benefited from a 3D measurement. The CCLZ survey to the south-west had less shallow complexity, but the basement was very deep. Both surveys had exploration objectives; there were no wells in the area and very limited seismic coverage. The primary objectives were to identify leads for further research, but additionally to improve basement imaging as an aid to obtaining a better basin model. In addition, CNOOC used the project as an opportunity to evaluate different broadband acquisition solutions. Acquisition configuration For both surveys, the vessel deployed a simultaneous over/ under source comprising two sub-arrays deployed at 6 m depth and two sub-arrays at 12 m depth (Figure 4). Firing of the deep sub-arrays was delayed such that the downgoing wavefields from the different depths positively interfered, and the upgoing wavefields that result in the source ghost destructively interfered. The survey also benefited from the use of a low-noise point-receiver acquisition system with full closed-loop streamer steering. More significantly, three streamers were deployed at depths of 5 m, 17 m, and 23 m (Figure 4). As well as allowing for the comparison of the three separate single depth images, the sparse-under (combination of 5 m and 23 m streamer depth datasets) and over/under (combination of 17 m and 23 m streamer depth datasets) results could also be evaluated. It should be noted that for this 2D experiment it is only possible to evaluate one component of the sparse-under solution, i.e., the use of a deep streamer measurement to boost the low frequencies recorded on a shallow streamer. We cannot infer anything about the Figure 3 Survey map showing the five pre-plot lines for the two survey areas. 2011 EAGE www.firstbreak.org 69

technical article first break volume 29, September 2011 Figure 4 Acquisition configuration with three streamers towed at different depths and the over/under source array. sparse cross-line sampling aspect of the 3D technique. In the following, we will refer to this subset of the complete sparse-under solution as low-frequency boosting. The potential improvement by applying each technique can, to some extent, be evaluated before acquisition by vertical-incidence modelling. In Figure 5 the amplitude responses of the three input wavelets (simultaneous over/ under source signature convolved with the three ghost responses) are shown on the left and of the two combined wavelets on the right. Clearly, the 5 m tow-depth wavelet has significantly attenuated low-frequency amplitudes (green curve). The 17 m and 23 m responses have significantly better low-frequency responses, but at the expense of ghost notches within the bandwidth of interest. The two combined responses compensate for the ghosting effect being rich in low frequencies without significant notching in the bandwidth of interest. This theoretical analysis suggests that the over/under summation wavelet has higher amplitudes at low frequencies as a consequence of summing the two deep-tow measurements which are both rich in low frequencies. At the higher frequencies the two solutions are very comparable and the minor differences are influenced by the choice of tow depths. Any analysis of this nature is imperfect because it does not include consideration of the noise response. The reality is that the improvement indicated by this signal-only analysis will not be fully realized; even if the two contributing noise responses are completely uncorrelated, we should expect a 3 db increase in the noise floor. For this analysis, in order to make the fairest representation, the high-frequency component of the low-frequency boosted wavelet does not include the summation component of the optimal de-ghosting process (6 db increase); the noise response would be fully correlated, and therefore there would be no improvement in the signal-to-noise ratio. Also, note that neither of the output wavelets benefit from the third step (spectral shaping/ weighting component) of the wavefield separation approach for similar reasons. Operational considerations It is fairly well accepted that over/under summation requires some level of streamer steering to ensure that the Figure 5 Amplitude spectra of vertical-incidence modelled wavelets for the three single streamers on the left, and the two combined wavelets on the right. 70 www.firstbreak.org 2011 EAGE

first break volume 29, September 2011 technical article Figure 6 Plot showing the lateral separations between the 5 m and 17 m depth streamers in blue and the 17 m and 23 m depth streamers in red. Figure 7 Amplitude spectra of shallow-streamer (blue) and deep-streamer (red) noise records with an absolute reference using the same spectral analysis window length. The green curve is the amplitude spectrum of the blended noise record. Figure 8 Intermediate image amplitude spectra from a 1-s window immediately below the water bottom. two streamers of the over/under pair measure the upgoing and downgoing wavefields at the same horizontal coordinates. One might assume that as the low-frequency boost solution limits the combination of the two towdepth measurements to low frequencies, it is less sensitive to lateral separation errors. While this might be intuitive, there is another factor to consider: the shallow streamer can be subjected to a very different current from the deeper streamers. A streamer towed between 5 m and 8 m depth is more affected by sea-surface effects. This is illustrated in Figure 6 which shows the tail-end lateral separation between the 5 m and 17 m streamers and that between the 17 m and 23 m streamers. Streamer-steering technology can compensate for the difference in current experienced by the streamers towed 17 m and 23 m below the sea surface (red curve). However, streamer steering cannot perfectly compensate for the very large current differential experienced by the streamers towed 5 m and 17 m below the sea surface, with lateral separation as high as 100 m in this example (blue curve), because the shallow streamer is influenced by very localized surface rip currents. Without Figure 9 Signal-to-noise ratio estimate from an intermediate image. streamer steering, the lateral separation could have been as high as 500 m at the tail end of the streamer, which would probably have been unacceptable even considering the lowfrequency limitations associated with the low-frequency boosting technique. Consequently, both over/under and sparse-under will benefit from the use of streamer steering to compensate for lateral separation errors caused by sea currents. The over/under and sparse-under solutions increase the signal-to-noise ratio by boosting signal levels at frequencies where the signal amplitude is very low, due to the ghost response. They also benefit from a further improvement in signal-to-noise ratio because they use deep streamers that are towed below the zone in which surface-related and weather-related noise is most significant. This makes both techniques less prone to acquisition interruptions due to bad weather periods. As the sparse-under solutions combine a deep-streamer measurement with a shallow streamer measurement, one might assume it does not benefit as much from this reduction in sea-surface-related noise. However, the sea-surface noise is generally at frequencies 2011 EAGE www.firstbreak.org 71

technical article first break volume 29, September 2011 Figure 10 Low-frequency filter panels (0 4 Hz bandwidth). Figure 11 High-frequency filter panels (32 64 Hz bandwidth). 72 www.firstbreak.org 2011 EAGE

first break volume 29, September 2011 technical article Figure 12 Comparison of heritage image (top) with final broadband images (low frequency boosted (middle) and over/ under (bottom)). The comparison details the full solution uplift including the broadband cable combination, the over/ under source and the low noise point receiver acquisition system but also demonstrates the very comparable uplift of the two broadband solutions. 2011 EAGE www.firstbreak.org 73

technical article first break volume 29, September 2011 below 20 Hz where the deep streamer data make the much greater contribution to the output. In simplistic terms, the combined low-frequency boosted noise floor is equivalent to the deep streamer noise floor below the blend frequency, where the surface-related noise resides, and equivalent to the shallow streamer noise floor above the blend frequency (green curve in Figure 7). Therefore, both combination techniques benefit from a reduction in noise floor, improvement in signal-to-noise ratio, and potential improvements in acquisition efficiency. Results To evaluate the benefits of the various datasets, spectra, filter panels, and migrated images were compared and evaluated. Analysis of spectra in various windows clearly demonstrates the value of the dual-tow depth combinations in extending the image bandwidth (Figure 8). The single tow-depth spectra illustrate the attenuation resulting from the receiver ghost. The low frequencies in the 5 m tow-depth streamer response (red curve) are heavily attenuated, whereas there are deep notches within the bandwidth of interest of the 23 m tow-depth streamer response (blue curve). Both over/under and lowfrequency-boosted responses are rich in high and low frequencies. It is also interesting to note that the data spectra are very comparable with the pre-survey 1D modelling (Figure 5). The goal of both dual-tow depth techniques is to increase the signal-to-noise ratio at the frequencies where it is poor due to the attenuation associated with the receiver ghost. Using an auto-correlation/cross-correlation technique, it is possible to estimate a separation of the signal (coherent) and noise (incoherent) responses in the final image (Figure 9). The two broadband techniques result in an increase of up to 10 db below 30 Hz, when compared to the 5 m single tow-depth measurement. This improvement can be further illustrated with lowfrequency filter panels (Figure 10). At these ultra-low frequencies (0-4Hz), the shallow-tow ghost response (left) is so aggressive that the signal levels of the data are very low. The two combined records (centre right and centre left) are similar and arguably a little better than the deep single-tow-depth record (right). However, if we analyse a higher frequency (32-64 Hz) panel (Figure 11), we see that the opposite is true. In this example, the single deep-tow depth measurement (right) has very poor signal-to-noise ratio. The two combined solutions (centre left and centre right) are comparable, but again better than the 5 m single tow-depth measurement (left). All products generated from this project will benefit from the low-frequency extension associated with the over/under source and the low noise floor associated with the single-sensor sampling and processing. All comparisons shown so far only evaluate the benefit of the streamer combination and do not evaluate the benefit of the broadband source or low-noise point-receiver acquisition system. One of the sequences was shot parallel and close to a heritage 2D line that had been recently re-processed using a similar processing flow. With the exception of tow depth, the acquisition parameters were very similar; in fact, the heritage line towed a slightly longer streamer. Comparison with this heritage line illustrates the full benefit of dual-depth sources and streamers to mitigate the band-limiting effect of the seasurface ghost (Figure 12). As well as significantly improving imaging of the deep basement (9 s two-way time), reflectors at shallow and mid depths are also much better imaged. Conclusions The presented field data clearly demonstrate the benefit of dual tow-depth solutions in both extending bandwidth and mitigating the impact of tow depth on the final image. This has benefits in extending the value of a single seismic experiment to multiple targets and will additionally result in a more stable input to seismic attribute analysis, including AVO inversion. The improvement is particularly apparent when the improvement associated with the full acquisition system is evaluated, i.e., simultaneous over/under source, dual-depth streamer combination, and a low-noise, point-receiver acquisition system. Both broadband solutions show comparable improvement. While the 1D modelling suggests that the over/under solution should provide an improved signal-to-noise ratio at the low frequencies compared to the low-frequency boosted solution, the improvement for this field data example is incremental, at best. The potential leverage of a sparsely sampled deep streamer plane to improve acquisition efficiency in conjunction with the low-frequency boosting solution makes it an attractive alternative for 3D surveys. Other operational issues associated with the use of a shallow streamer have a much smaller impact on operational efficiency. Acknowledgements The authors thank CNOOC (CNOOC-Zhanjiang and CNOOC Shenzhen) for their willingness to share the results and the many contributions from the WesternGeco technical team. References Hill, D., Combee, L. and Bacon, J. [2006] Over/under acquisition and data processing: the next quantum leap in seismic technology? First Break, 24(6), 81 96. Kragh, E., Svendsen, M., Kapedia, D., Busanello, G., Goto, R., Morgan, G., Muyzert, E. and Curtis, T. [2009] A method for efficient broadband marine acquisition and processing. 71 st EAGE Conference and Exhibition, Extended Abstracts, V018. Özdemir, A.K., Caprioli, P., Özbek, A., Kragh, E. and Robertsson, J.O.A. [2008] Optimized deghosting of over/under towed-streamer data in the presence of noise. The Leading Edge, 27, 190 199. Posthumus, B.J. [1993] Deghosting using a twin streamer configuration. Geophysical Prospecting, 41, 267 286. Received 5 January 2011; accepted 5 July 2011. doi: 10.3997/1365-2397.2011027 74 www.firstbreak.org 2011 EAGE