Nick Moldoveanu, WesternGeco Summary Marine seismic data have been traditionally contaminated by bulge waves propagating along the streamers that were generated by tugging and strumming from the vessel, paravanes, tail buoys, and lead-in cables. With the progress of streamer technology bulge-wave interference has been significantly reduced. However, weather and flow noise still affects marine seismic data. The level of cross-flowinduced noise is increased when the data are acquired during turns or along circles, like in Coil shooting, and when marine currents are strong. In this paper, we present a new technique to attenuate towed-streamer noise acquired in these conditions. The method has been used successfully to process coil and wide-azimuth data in the Gulf of Mexico and offshore Brazil. Coil shooting acquisition acquires data along circles that overlaps in x- and y-directions to cover the entire survey area (Moldoveanu, 2008). For coil shooting, the level of the horizontal cross-flow noise vs. the low-frequency signal can be higher than 20 db when ocean currents affect the streamer spread. Figure 2 shows an example of a raw shot gather recorded during coil shooting acquisition with a point-receiver streamer that has single hydrophones spaced at 3.125-m intervals, and without any acquisition filter. The corresponding FK spectrum and amplitude spectrum are displayed in Figures 3 and 4. It can be seen that the noise is 35 db stronger than the signal at low frequencies. Introduction Towed-streamer marine acquisition technology evolved significantly in the last decade in terms of the in-sea equipment, particularly streamers, streamer control devices and towing systems, and the result of this was a reduction of noise induced by tugging, birds (streamer control devices) and electrical interferences. However, towed marine data are still affected by the vertical and horizontal cross-flow of water across the streamers. Vertical crossflow can be induced by wave action and results in so-called swell noise. Horizontal cross-flow is induced by ocean currents and as the vessel turns or when the vessel sails along a circular path. All sources of cross-flow generate vibrations that propagate along the streamer and are recorded as high-amplitude low-frequency noise. Figure 1 (Curtis and Davis, 2001) shows typical signal and ambient noise spectra for towed streamers prior to array forming. The cross-flow noise is more than 20 db higher in amplitude than the seismic signal in the frequency range of 0 to 5 Hz and comparable with the signal from 5 to 10 Hz. Figure 2: Raw point-receiver gather acquired during a coil shooting survey Figure 3: FK-spectrum of the raw point-receiver record Figure 1: Signal (blue) and ambient noise (red) for towed streamers prior to array forming Improving the signal-to-noise ratio in the low-frequency range is important for imaging deep targets, velocity model building and seismic inversion. The latest developments in SEG San Antonio 2011 Annual Meeting 3576
velocity model building using full-waveform inversion (FWI) require very low frequencies, in the range 0 to 6 Hz, to obtain maximum resolution of the velocity field (Vigh et al., 2010). Increasing the signal at the very low frequencies could be an alternative to improve the signal-to-noise ratio, but this is not practical with the airgun source technology available today. For this reason it is important to develop effective noise attenuation algorithms for low-frequency marine noise. Figure 4: Amplitude spectrum of the raw point-receiver record Ozbek (2000) introduced the linearly-constrained adaptive noise attenuation (LACONA) method that has been used successfully as a core component of noise attenuation workflows to attenuate the swell noise on marine seismic data recorded with finely sampled point receivers (Martin et al., 2000). The method proposed in this paper adds a preconditioning step to such workflows that addresses the strong horizontal cross-flow noise recorded during vessel turn or coil shooting acquisition. The method presented here is based on the following assumptions: Data=Seismic + Noise Noise amplitudes >> Signal amplitudes Largest singular values of the matrix D correspond to the largest amplitude values, which are associated with the cross-flow streamer noise In our application the matrix D corresponds to a shot gather or a sub-gather (group of traces) that has m samples and n traces. If the singular values of matrix D are calculated and sorted in decreasing order, can select the largest k singular values s s... 11 22 nn, we s 11, s22,... skk, and reconstruct a matrix N U1 * S1 * V. N represents an estimation of the noise and has the same dimension, (m,n), as the data matrix D. This allows subtraction of the noise from the data, S 1 D N, where S1 is a representation of the seismic signal plus residual noise. The noise is estimated iteratively as shown in Figure 3. Noise estimation is done in a frequency band, typically 0 to 5 Hz or 0 to 10 Hz. The number k of largest singular values that will be kept in the SVD decomposition and the number of iterations are the critical parameters for this method. If these numbers are too high, the signal could be attenuated. In this implementation, the numbers of singular values and the number of iterations can vary from shot to shot as a function of the noise level. Also, the process can be stopped if the difference of the noise estimated in two consecutive iterations is less than a user-defined threshold. ' 1 s Method description and implementation Singular value decomposition (SVD) is well known in linear algebra and allows us to decompose a matrix D(m,n) with m rows and n columns, in a product of three unitary ' matrices, D U * S * V. Matrixes U, V and S have dimensions (m,n), (n,m) and (n,n), respectively. S is a diagonal matrix whose elements are the singular values of the matrix D. The SVD method has been used in seismic data processing for signal-to-noise ratio enhancement using Karhunen- Loeve transform (Jones and Levy, 1987), footprint removal (Al-Bannagi, 2005), and ground-roll attenuation. Two recent papers addressing SVD for ground-roll attenuation are Chiu and Howell (2008) and Cary and Zhang (2009). Figure 5: Iterative estimation of the noise using SVD method The criterion to detect the noisy traces is based on the calculation of the RMS amplitude in a window where the noise dominates the signal. SEG San Antonio 2011 Annual Meeting 3577
Data examples The examples included in the abstract are from a 2x4 coil shooting survey and a wide-azimuth (WAZ) survey acquired in the Gulf of Mexico in 2010 and 2011. Both surveys were acquired with a point-receiver system. Figure 6 shows the result after the SVD noise attenuation method was on the raw point-receiver gather shown in Figure 2. The FK spectrum derived after SVD noise attenuation is displayed in Figure 7 and the noise removed is shown in Figure 8. This example illustrates that the strong horizontal cross-flow noise recorded during vessel turns or during coil shooting acquisition can be efficiently attenuated using the SVD method, without affecting the underlying signal. The next example is from a WAZ survey. Figure 9 shows a point-receiver shot gather with swell noise. The data were recorded in rough weather conditions. A 1.75-Hz low-cut Kaiser filter was to this shot before SVD. The point-receiver shot gather after SVD is displayed in Figure 10, and the noise removed from the data is shown in Figure 11. This example again demonstrates that strong swellinduced cross-flow noise can be attenuated without damaging the low-frequency signal. Figure 8: Noise removed by the SVD method Figure 9: Point-receiver shot gather recorded in rough weather during a WAZ survey. 1.75 Hz low cut filter was Figure 6: Point-receiver gather after SVD noise attenuation was Figure 10: Point-receiver data after SVD noise attenuation was Figure 7: Point-receiver gather after SVD noise attenuation was SEG San Antonio 2011 Annual Meeting 3578
The SVD noise attenuation method discriminates the signal from noise based on amplitudes and requires careful testing to properly select the parameters that will protect the signal. Considering that the noise amplitudes are more than 30 db higher than the signal it is safe to attenuate the very highamplitude noise components. The rest of the marine noise is efficiently attenuated by a standard Lacona based noise attenuation workflow. Figure 13 shows an example of applying such a standard noise attenuation workflow on a point-receiver shot record processed through SVD (Figure 6). Figure 11: Noise removed by the SVD method Discussions and conclusions Our strategy for attenuation of the strong marine noise generated by cross-flow of water across the streamer is to to take advantage of recording single hydophone data with fine receiver sampling and no acquisition filter, and to have a multistep approach for noise attenuation in data processing. The processing seqeuence performed onboard of the vessel is shown in Figure 12. Figure 13: Result of standard noise attenuation workflow after SVD on the first data set The noise attenuation flow presented here is done in the shot domain, where the receiver sampling is 3.125 m. Although the SVD process is not sensitive to aliasing, the fine spatial sampling is required prior to further noise attenuation and receiver motion correction. The proposed flow using the SVD noise attenuation method was used to process 2x4 coil shooting data acquired in the Gulf of Mexico in 2010, a single vessel coil survey acquired ofshore Brazil and WAZ data acquired in the Gulf of Mexico in 2011. Noise attenuation methods based on SVD were used in the past to attenuate ground roll, random noise and acquisition footprints. We demonstrated that an iterative method based on SVD can be used to efficiently attenuate the high-energy streamer noise recorded during turns or coil shooting acquisition and strong swell noise. Acknowledgements I aknowledge WesternGeco for permission to present the paper and my colleagues Stephen Bracken and Kristen Doty for their contribution to the implementation and testing of this new method. Figure 12: Onboard processing for strong marine noise attenuation SEG San Antonio 2011 Annual Meeting 3579
EDITED REFERENCES Note: This reference list is a copy-edited version of the reference list submitted by the author. Reference lists for the 2011 SEG Technical Program Expanded Abstracts have been copy edited so that references provided with the online metadata for each paper will achieve a high degree of linking to cited sources that appear on the Web. REFERENCES Al-Banngi, M., K. Fang, P. G. Kelamis, and G. S. Douglass, 2005, Acquisition footprint suppression via the truncated SVD technique: Case studies from Saudi Arabia: The Leading Edge, 24, 832 834, doi:10.1190/1.2032259. Cary, P., and C. Zhang, 2009, Ground roll attenuation via eigenimage filtering: 79th Annual International Meeting, SEG, Expanded Abstracts, 3302 3305. Chiu, S. K., and J. E. Howell, 2008, Attenuation of coherent noise using localized-adaptive eigenimage filter: 78th Annual International Meeting, SEG, Expanded Abstracts, 2541 2545. Curtis, T., and T. Davis, 2001, Extending the bandwidth of marine data: 71st International Exposition and Annual Meeting, SEG, Expanded Abstracts, pp. 37 40. Jones, I. F., and S. Levy, 1987, Signal-to-noise ratio enhancement in multichannel seismic data via Karhunen-Loeve transform: Geophysical Prospecting, 35, no. 1, 12 32, doi:10.1111/j.1365-2478.1987.tb00800.x. Martin, J., A. Ozbek, L. Combee, N. Lunde, S. Bitleston, and E. Kragh, 2000, Acqusition of point receiver seismic data with a towed streamer: 70th Annual International Meeting, SEG, Expanded Abstracts, 37-40. Moldoveanu, N., 2008, Circular geometry for wide-azimuth towed streamer surveys: 70th Annual International Conference and Exhibition, EAGE, Extended Abstracts, 55 59. Ozbek, A., 2000, Adaptive beamforming with generalized linear constraints: 70th Annual International Meeting, SEG, Expanded Abstracts, 2081 2084. Vigh, D., B. Starr, J. Kapoor, and H. Li, 2010, 3D full waveform inversion on a Gulf of Mexico WAZ data set: 80th Annual International Meeting, SEG, Expanded Abstracts, 957 960. SEG San Antonio 2011 Annual Meeting 3580