A Publication of Petroleum Geo-Services Vol. 9 No. 7 July 2009 Conventional vs. GeoStreamer Data De-ghosting: The Haltenbanken Dual Streamer Case study Introduction The Haltenbanken dual streamer data was acquired in 2007 on the mid-norwegian continental shelf in the Norwegian Sea. The objectives are at various prospective target depths in a deep and relatively complex structural and stratigraphic geological setting. Major hydrocarbon discoveries are located in pockets of Jurassic sediments, mostly on the Halten terrace. These Jurassic sediments are delimited by tilted fault blocks, and capped by Cretaceous cap-rocks. There are also reservoir-quality sandstones Summary We compare conventional data with dualsensor (GeoStreamer ) data acquired simultaneously at two different depths. The data were acquired using a common source, and recorded by both a dual-sensor streamer (GeoStreamer) at 15 m depth and a conventional streamer at 8 m depth. To facilitate comparison, the data from the two streamers have been processed optimally and in the same manner, i.e. wavefield decomposition was also conducted on the conventional streamer data. We compare the resulting conventional up-going pressure field to the up-going pressure field derived from the GeoStreamer data. The comparison shows a clear advantage for the dual-sensor wave field decomposition result. The superior signal-to-noise ratio from the GeoStreamer data is attributed to the deeper towing depth and more accurate deghosting. No assumptions were made about the sea surface state when processing the GeoStreamer data. within the Cretaceous that can constitute prospective targets. The challenge is therefore to image sediments and fault planes as precisely as possible, which means acquiring high resolution seismic data. We also need good signal penetration as some of the structures of interest have more than 10 seconds two-way travel time (TWT). Figure 1: Location of the Haltenbanken survey. The 2D acquisition geometry comprised two spatially coincident streamers, each 8100 m long, towed in the Continued on next page
TechLink July 2009 Page 2 Introduction Continued from Page 1 same vertical plane. A conventional hydrophone-only streamer was towed at 8 m depth below sea-level, whereas a dual-sensor streamer (GeoStreamer ) was towed below it at 15 m depth. The common source array was towed at 7 m depth. Two lines of data were acquired using this geometry on the same prospect. The location of the two lines ( Line 1 and Line 2 ) is indicated in Figure 1. This dataset presents an opportunity to compare the images obtained using the GeoStreamer with that obtained from a conventional hydrophone-only streamer towed in an acquisition configuration typical of North Sea surveys. The results discussed below are from Line 1, which was acquired in varied weather conditions (marginal at the start of the line, calming out towards the end of the line). A dual-sensor streamer records both pressure and the vertical component of particle velocity with collocated sensors, allowing for very efficient decomposition of the total pressure wavefield into up- and down-going components at the receiver locations. This operation is often called (receiver-side) de-ghosting. This procedure and the consequent data quality enhancements have been demonstrated in several publications (e.g. Carlson et al., 2007; Long et al., 2008). The technique used for deghosting dual streamer data was developed specifically for dual sensor acquisition. A natural question is whether these same techniques can be applied to conventional data (cf. Amundsen, 1993). We address that question below in the comparison of datasets acquired using the two different streamer types. Pre-processing and Wavefield Decomposition The theory behind wavefield decomposition has been described by a number of authors, e.g. Amundsen (1993) and Fokkema and van den Berg (1993). A brief description of the pre-processing steps that allow the decomposition of data into up- and down-going components at the receiver locations is presented below. When the vertical component of the particle velocity wave-field is measured in a spatially-coincident manner with the total pressure wavefield, the two components can be used directly to decompose the total pressure and particle wavefields into their up- and down-going components. For the pressure wave-field the two components are given as: (1) P is the measured total pressure wavefield, V z is the measured vertical component of the total particle velocity wavefield, and F is an angle-dependent scaling factor. In the frequency wave-number domain, the scaling filter F is (Amundsen, 1993): (2) where k x, k y and k z denote the three components of the angular wave-number vector, ω denotes angular frequency, and ρ and v w are the density of water and the acoustic wave propagation velocity in water, respectively. The scaling factor includes an angle-dependence that is required because we record only the vertical component of the particle velocity vector. Note that no assumptions are made about the sea surface state. In practice, the particle velocity data is often heavily contaminated with low-frequency noise. To circumvent this problem, the lowest frequencies are rebuilt from the pressure record using the frequency-wave-number solution of the equation of motion (Tenghamn et al., 2007): (3) where z is the receiver depth, and r0 is the magnitude of the reflection coefficient. This process assumes a flat sea surface with a constant reflection coefficient. For all the examples in this paper, a reflection coefficient of -1 has been assumed. The range of frequencies for which the recorded particle velocity data is rebuilt using equation (3) is typically limited to 20 Hz and lower, though it is anticipated that future technological advances will allow this limit to be lowered even further. Equation (3) constitutes the basis of a method for deghosting conventional hydrophone-only data. By substituting the expression for V z from equation (3) into equation (1), one can in theory decompose conventional hydrophone-only data into up- and down-going components. The only mathematical limitation is that equation (3) cannot be used (as it is singular) at the ghostnotch frequencies of the hydrophone data. Hence, in a practical application, the method may only be used up to the second notch in the amplitude spectrum of hydrophone data. Furthermore, unlike the dual-sensor case, the flat sea surface assumption must hold at all frequencies. The
Page 3 A Publication of Petroleum Geo-Services validity of this assumption at different frequencies is considered below. Processing and Results Identical processing flows were applied to data acquired by each of the two streamers. The main processing steps include noise-suppression in both x-t and τ-p domains, water-bottom gapped deconvolution on common receiver gathers after minimum phase conversion, surface-related multiple elimination (SRME), and finally Kirchhoff pre-stack time migration. The only difference in the processing of the GeoStreamer data and the conventional hydrophone-only data is in the wavefield separation step. The latter was performed at the same location in the corresponding processing flows directly after the noise suppression step. After wavefield separation, the up-going pressure field from the dual-sensor GeoStreamer was re-datumed from the 15 m acquisition depth to 8 m depth before further processing to facilitate direct comparison with the conventional streamer data. Note that this re-datuming process only affects travel times, and not the amplitude spectrum at any TWT. Figure 2 shows a comparison between the final migrated stack of the conventional total pressure (hydrophone) data without wave field decomposition and the same stack for the upgoing pressure field obtained from the dual-sensor GeoStreamer data. Note the superior illumination and resolution of the processed GeoStreamer data. Even though the GeoStreamer data shows superior imaging capabilities, it remains difficult to draw clear conclusions, as we are attempting to Figure 2: Top: Final migrated image using the total pressure wavefield recorded by the conventional streamer at 8 m depth. A deep window containing several compare two seismic sections with completely intermediate-to-deep targets and a shallow window are highlighted. Bottom: different signal characteristics. The section derived Equivalent image obtained from the up-going pressure wavefield derived from the dual-sensor GeoStreamer at 15 m depth. Comparative mean spectra for the from GeoStreamer data contains de-ghosted deep window are also shown. signal, i.e. only the up-going pressure wave-field, signal characteristics of the two datasets deterministically whereas the other section from the conventional by de-ghosting the conventional pressure-only data using a hydrophone data contains both the primary (up-going combination of equations (3) and (1) as described above. signal) and its ghost reflected from the sea-surface (downgoing signal). In order to better compare these datasets, We chose to match the conventional data to the GeoStreamer data, rather than the reverse, since the deghosted data is the most desirable output. De-ghosting we ideally need to match the signal characteristics of the data derived from the two streamers. We matched the Continued on next page
TechLink July 2009 Page 4 Processing and Results Continued from Page 3 leads to a shortening of the wavelet and consequently enhanced resolution. Deterministic signal matching is preferred to statistical signal matching, given that we would normally acquire either conventional hydrophoneonly data or GeoStreamer streamer data, and would be tasked with producing the best possible image from that data in isolation. Hence, this experiment with both streamer types represents a unique opportunity to assess the viability of de-ghosting conventional streamer data and the resulting image quality compared with that obtained from GeoStreamer data. Figure 3 shows detail of migrated stacks for some of the intermediate to deep prospective targets highlighted in Figure 2, but where both the conventional and GeoStreamer data have been deghosted. This figure shows that the superior image quality observed in Figure 2 is maintained after deghosting of both datasets. Inspection of this figure also shows that the signal characteristics and amplitudes obtained from the two streamers are more similar after de-ghosting of the conventional streamer data. This similarity is reflected by the amplitude spectra of the two sections, where discrepancies are observed only at very low and very high frequencies. Note that, for the conventional streamer, we have applied an 80-88 Hz low-pass filter. This is necessary because the de-ghosting process is restricted by the second hydrophone notch, which is at ~94Hz for an 8 m tow depth, as shown in Figure 4. No such limitation applies to the GeoStreamer for which deghosting is achieved by combination of data from the two sensors. As the signal characteristics of the two sections in Figure 3 have been matched, the differences in the migrated stacks are caused by differences in the noise recorded by the two streamers. This is particularly evident in the deep part of the section, where many reflectors that are observed on the GeoSstreamers seismic section are harder to identify, or entirely obscured by noise on the conventional streamer section. Figure 3: Top: Final migrated image using the up-going pressure wavefield derived from the conventional streamer at 8 m depth for the deep window in Figure 2. Bottom: Equivalent image obtained from the up-going pressure wavefield derived from the GeoStreamer at 15 m depth. Comparative mean spectra are also shown. Whilst the gross characteristics of the signal are observed to be similar for the two streamers after de-ghosting, inspection of the shallow data in Figure 5 reveals that there are important differences in the detail. These shallow events are much better resolved for the GeoStreamer data than for the conventional data. These differences in signal characteristics imply that the de-ghosting of the conventional data has been less successful than for the GeoStreamer data. Recall that when
Page 5 A Publication of Petroleum Geo-Services de-ghosting the conventional streamer the flat sea surface assumption must be used for the entire bandwidth, whereas in the case of the dual-sensor GeoStreamer this assumption is required only for the lowest frequencies. The higher frequencies are relatively more important for imaging the shallow data shown in Figure 5 than they are for the deeper data. Hence, it is suggested that the primary cause of the differences in the shallow images derives from the inappropriate use of a flat sea surface assumption when de-ghosting the conventional streamer data. This theory will be tested further in the following section. Discussion It has been shown that the image quality obtained from the dual-sensor GeoStreamer is superior to that obtained from the conventional streamer at all depths after de-ghosting to deterministically match the signal characteristics of the data from the two streamers. The primary reasons for the discrepancies in image quality have been identified as differences in signal-to-noise of the two datasets, and the assumptions necessary to de-ghost the conventional streamer data. These issues are discussed further in this section. For the very lowest frequencies, de-ghosted data are derived solely from the pressure sensor for both the conventional and dual-sensor streamer (GeoStreamer). The effect of receiver-side deghosting is the inverse of the ghost spectra shown in Figure 4 for 8 m and 15 m tow depths, such that after de-ghosting, the signal from both recording depths will be equalized. However, the de-ghosting operator also acts on any noise present in the field Figure 5: Top: Final migrated image using the up-going pressure wavefield derived from the conventional streamer at 8 m depth for the shallow window in Figure 2. Bottom: Equivalent image obtained from the up-going pressure wavefield derived from the GeoStreamer at 15 m depth. Comparative mean spectra are also shown. Figure 4: Theoretical amplitude spectra of the hydrophone ghost function for 8 and 15 m tow depth. data. From inspection of Figure 4, the shape of the amplitude spectra clearly shows that even if the noise level is identical at the two recording depths, the 15 m tow depth will provide higher signal-to-noise ratio than the 8 m tow depth at the lowest frequencies. Hence, a higher level of low frequency noise is expected for the 8 m deep streamer after de-ghosting, solely due to the relative shape of the ghost functions in the vicinity of the notch at 0 Hz. Continued on next page
TechLink July 2009 Page 6 Discussion Continued from Page 5 In practice, the conventional streamer is subject to more weather-related noise in marginal weather, due to its shallower tow depth of 8 m compared to the relatively quieter conditions encountered at the 15 depth at which the GeoStreamer is towed. This effect is illustrated in Figure 6, where average amplitude spectra calculated from start-of-line noise records are compared. The noise records were acquired simultaneously on the two streamers at the beginning of Line 1 under marginal weather conditions. Towing deeper under such conditions is clearly advantageous. The lower noise level, combined with the enhanced signal illustrated in Figure 4, means that the signal-to-noise ratio at 15 m tow depth is significantly higher than at 8 m for low frequencies. This difference in signal-to-noise content accounts for the superior imaging of deep targets obtained from the GeoStreamer, even though the frequency content at depth is such that the data are derived solely from the pressure sensor in both cases. The effect of the low frequency noise is readily apparent from inspection of the comparative sections in Figure 3. Note that a 1-3 Hz high-pass filter has been applied to all images in this analysis. This filter is routinely applied to mitigate the worst effects of weather-related noise on image quality. The spectral comparison in Figure 3 shows higher amplitudes at the very lowest frequencies. Given Figure 7: Top: Final migrated image for the deep window in Figure 2 using the up-going pressure wavefield derived from the conventional streamer at 8 m depth after application of a 2-5 Hz high-pass filter. Bottom: Equivalent image obtained from the up-going pressure wavefield derived from the GeoStreamer at 15 m depth (no further filtering applied). Comparative mean spectra are also shown. Figure 6: Mean amplitude spectra of noise recorded at the start of acquisition by the conventional streamer at 8 m depth and the pressure sensor of the dual-sensor GeoStreamer at 15 m depth. that we expect there to be very little signal below 5 Hz for the conventional streamer, this amplitude discrepancy is caused by enhancement of weather-related noise by the de-ghosting operator. Hence, in this comparison the conventional streamer section is arguably disadvantaged by the presence of extra noise at frequencies that do not
Page 7 A Publication of Petroleum Geo-Services greatly contribute to the image quality. Figure 7 shows the conventional streamer section from Figure 3 after application of a 2-5 Hz low-cut filter. The amplitude spectrum is much closer to that derived from the dualsensor streamer at low frequencies after the application of this filter, and the image is correspondingly cleaner. However, it is also evident that the imaging of the deep reflectors is still inferior to that obtained from the GeoStreamer. The structural interpretation that would be obtained from the conventional streamer section has not been enhanced by the filtering operation. If a conventional hydrophone-only streamer was towed at 15 m depth to take advantage of the quieter recording conditions, a similar image of the deep targets would be obtained. However, the usable seismic bandwidth would be restricted to below the second hydrophone notch at about 50 Hz. This would have an adverse impact on resolution, especially for the shallower targets. By using the GeoStreamer, data at the hydrophone notch frequencies is provided by the complementary signal recorded from the particle velocity sensor. superposition of three sinusoids with similar wavelengths. The dominant wavelength was 120 m and the wave height (peak to mean sea level) was 2 m, representing a rough sea state close to the limit that seismic data can safely be acquired. A source with a signature typical of a marine seismic source was placed at depth to simulate the signature that would be observed in the far field. Figure 8 shows modeling results for 15 m and 8 m receiver depths. This figure shows superimposed amplitude spectra of the reference signal (i.e. flat sea) and three other modeled signals from different locations beneath the simulated rough sea surface. Note the progressive divergence of the spectra as frequency increases. When de-ghosting the conventional streamer towed at 8 m depth, the flat sea surface assumption must be used up to 90 Hz. It is clear from Figure 8 that the flat sea surface assumption breaks down in these circumstances. However, for the Continued on next page By towing the conventional streamer at 8 m depth, which is typical of surveys in the North Sea, we may perform de-ghosting up to ~90 Hz in order to obtain signal characteristics that are comparable with those obtained from the GeoStreamer. In order to perform this procedure, we assume a flat seasurface with constant reflection coefficient for the full bandwidth. The same assumptions are made when processing GeoStreamer data, but only for the lowest frequencies. Both assumptions will be violated when the sea-surface shape is not flat. The amplitude and phase of the ghost reflections will vary in a complicated and time-variant manner depending on the wave heights and wave lengths above the receiver locations. Methods for taking account of rough sea surfaces in processing require information about the sea surface shape that is not readily available (see, for example, Robertsson and Kragh, 2002, and Amundsen et al., 2005). In order to investigate the effects of a rough sea surface, a simple finite difference modeling study was conducted. Receiver arrays were located below a simulated rough sea surface created by the Figure 8: Comparative theoretical amplitude spectra for a flat sea surface (red) and three locations beneath a rough sea surface with 2 m wave amplitude and 120 m dominant wavelength. Results are shown for 8 m (top) and 15 m (bottom) hydrophone depths. The dashed vertical lines denote the frequencies up to which pressure-only de-ghosting must be performed for the conventional and dual-sensor streamer cases respectively.
TechLink July 2009 Page 8 Discussion Continued from Page 7 GeoStreamer towed at 15 m depth, the flat sea surface assumption clearly holds within the limited range of 0 20 Hz required for GeoStreamer de-ghosting. These results are consistent with the images shown in Figure 5, which show superior resolution for the GeoStreamer. This is attributed to the use of recorded particle velocity data over the most of the bandwidth, and the restriction of the flat sea surface assumption to only that part of the data for which this assumption is valid. Conclusions We have compared the final image quality for data acquired using a conventional hydrophone-only streamer towed at 8 m depth to data contemporaneously acquired with the dual-sensor GeoStreamer towed below at 15 m depth. To facilitate this comparison, the conventional data were de-ghosted up to the second hydrophone notch frequency, in order to provide a signal that is directly comparable with the up-going pressure field obtained from standard dual-sensor streamer processing of the GeoStreamer. For the deeper targets, the GeoStreamer quite obviously has superior signal-to-noise content. This benefit follows from the greater towing depth, due to the comparative shape of the ghost function for different depths, and also because the GeoStreamer is subjected to reduced levels of weather-related noise. Towing a conventional hydrophone-only streamer at a similar depth is not a viable proposition without severely compromising the usable seismic bandwidth, and consequently, resolution. For a conventional towing depth, a flat sea surface CONTACT Petroleum Geo-Services London Tel: +44 1932 376000 Fax: +44 1932 376100 Oslo Tel: +47 67 52 6400 Fax: +47 67 52 6464 Houston Tel: +1 281 509 8000 Fax: +1 281 509 8500 Singapore Tel: +65 6735 6411 Fax: +65 6735 6413 2009 Petroleum Geo-Services. All Rights Reserved assumption can be used to de-ghost conventional streamer data. However, this assumption breaks down at higher frequencies, with consequent degradation of image resolution. Hence, the dual-sensor GeoStreamer in fact provides a clearly superior image quality at all depths. Acknowledgements We thank the crew of Ramform Explorer who acquired the data shown here, and many colleagues within PGS for their assistance and contributions. We also thank PGS for permission to publish this data. References Amundsen, L., 1993, Wavenumber-based filtering of marine point-source data. Geophysics, 58, 1335-1348. Amundsen, L., Røsten, T, Robertsson, J., and Kragh, E., 2005, Rough-sea deghosting of streamer seismic data using pressure gradient approximations. Geophysics, 70, V1-V9. Carlson, D., Long, A., Söllner, W., Tabti, H., Tenghamn, R., and Lunde, N., 2007, Increased resolution and penetration from a towed dual-sensor streamer. First Break, 25, 12, 71-77. Fokkema, J.T., and van den Berg, P.M., 1993, Seismic applications of acoustic reciprocity. Elsevier, Amsterdam, 350p. Long, A., Mellors, D., Allen, T., and McIntyre, A., 2008, A calibrated dual-sensor streamer investigation of deep target signal resolution and penetration on the NW Shelf of Australia. SEG Expanded Abstract 27, 428-432. Robertsson, J. and Kragh, E., 2002, Rough-sea deghosting using a single streamer and a pressure gradient approximation. Geophysics, 67, 2005-2011. Tenghamn, R., Vaage, S., and Borresen, C., 2007, A dualsensor towed marine streamer; its viable implementation and initial results. 77th Annual International Meeting, SEG, Expanded Abstract, 989-993. For Updates on PGS Technological Advances, visit www.pgs.com More TechLinks at www.pgs.com/techlink