Summary The ghost delay time is a function of the ray parameter. In the case of 3D data in the shallow water environment, the crossline component of the ray parameter is not negligible. Conventional interpolation techniques combined with the use of a high-resolution Radon transformation help reduce the aliasing effect. This combination facilitates a full 3D deghosting process, in which both inline and crossline components of the horizontal slowness are acknowledged. Introduction In recent years, the deghosting of conventional 2D marine seismic data has become a standard process.
The ghosting effect of towed marine seismic data is controlled by the acquisition geometry and the sea state. Deterministic methods of deghosting typically require accurate depth information for every receiver along the length of the streamer within decimetres. Any minor inaccuracy in this information can lead to characteristic ringing through application of the deghosting operator in the wrong frequency. In practice neither the sea surface is flat nor do the receivers remain at their nominal depths; measurements themselves are sparser and generally interpolated. The position of the receiver-ghost notch frequency is dynamic, varying for every receiver in every shot gather which is augmented in higher sea states.
Here we describe an approach of differential deghosting applied to a 3D dataset offshore west Africa. Firstly, the receiver ghost notch is isolated from the f-x spectra of the precritical water bottom reflection for every shot and a search of minimum amplitude is performed around the calculated value from the recorded measurements. Based upon these estimates we move or ‘reghost’ the notch either to the measured value in the trace headers or to a corresponding nominal depth. By applying this step we demonstrate that the variability in the sea surface is accounted for and a significant improvement made in subsequent full deghosting.
Deterministic methods for deghosting marine seismic data assume we know both the source and receiver depths, reflection coefficients and account for the ghost variation with incident angle. If the sea surface is flat acting as a ‘perfect mirror’ these assumptions can be valid but rarely this is true. As shown in Figure 1, the position of the receiver ghost notch is dynamic from shot to shot and along the streamer itself. This high level of dynamicity results in the random diversification of the receiver ghost notch frequency for any given angle of incidence.
By applying deterministic deghosting only we can recover the amplitude within the receiver ghost notch for any given angle prestack, yet a phase discrepancy may still occur from receiver to receiver. This may only become apparent when the data is stacked, manifesting itself as a ‘residual notch’. The cartoon in Figure 2 demonstrates this effect whereby the streamer itself is generally well behaved but the sea surface is varying. Only by compensating for this variation can all events be summed constructively to avoid a residual receiver ghost notch effect post stack (Figure 3).
The frequency of ghost notch is naturally diversified by random variations at the sea surface. Further diversity may be achieved by towing a variable depth streamer. Regardless of the streamer shape however, the recorded seismic data needs redatuming and deghosting with respect to both ray parameter and individual source and receiver depths. We present a slowness-variant redatuming and deghosting method, capable of handling variable depths and irregular offsets. We utilize fine-tuned depths for deghosting and estimated elevations for redatuming. The effectiveness of the presented method has been validated by application to field data acquired by various configurations including flat, slant and curved streamers.
Broadband seismic data is desirable for interpretation purposes. In a marine environment however, the spectrum of recorded seismic signal is governed by a number of factors including the interference of the ghost reflections. Attempts are made to improve both temporal and spatial resolution of seismic data in both acquisition and processing stages. Recent developments in the acquisition stage include dual-sensor streamers (Carlson et al., 2007), over-under streamers (Özdemir et al., 2008), variable-depth streamers (Soubaras, 2010) and multicomponent streamers measuring pressure and the gradient wavefields (Vassallo et al., 2013). In the processing stage a number of techniques have been introduced aiming to deghost the data either before, while, or after the migration process. A premigration deghosting method was proposed by Wang and Peng (2012). Zhou et al. (2012) applied a deghosting process on conventional streamer data. Masoomzadeh et al. (2013) proposed a method of angle-dependant redatuming and deghosting for slanted streamers, assuming a flat sea surface. Robertsson and Amundsen (2014) developed a finite-difference method for deghosting streamer data towed at arbitrary variable depths. In this paper we propose a method of redatuming and deghosting applicable to seismic data acquired by single-sensor variable-depth streamers in presence of sea surface undulations.
In practice, neither the sea surface is flat nor do the receivers remain in predefined depths. As can be seen in Figure 1, the receiver depths are constantly changing, even during the listening period. This level of variation results in a ‘natural’ diversity of nonzero notch frequencies. This nearly random diversity means that the weak signals surviving the destructive ghost interference in the vicinity of the nominal notch frequencies can be further suppressed by the stacking process.
In 2013 TGS added additional coverage to its 3D database in the Norwegian Barents Sea over the Hoop and Finnmark Platform areas. Significant benefits are observed from the broadband processing of both new and previously acquired data through accurate deterministic deghosting and a deconvolution process, stabilized using large statistics. Sea conditions in the Barents Sea are rarely benign which can lead to unintentional variations in the streamer shape and depth. These variations can pose challenges for deterministic deghosting if the streamer is assumed to be flat and the sea surface reflection coefficient constant, as both are dynamic. A strategy is described to cope with such variations by combining deghosting techniques for flat and slanted streamer acquisition into a hybrid approach using examples from the Hoop area. In the Finnmark Platform area we correlate the ambient noise and streamer variation to observations on the sea state. By assigning the data into different quartiles the sea surface reflection coefficient can be estimated and varied as a seed for a stochastic search to find the most appropriate set of deghosting parameters.
Fault shadows represent zones of unreliable seismic imaging in the footwall of major extensional faults. They occur due to rapid lateral velocity variations and high velocity overburden which distort raypaths, manifesting themselves as sags on a time image. While interpretation driven techniques in the depth domain have been developed to address this issue, the problem is fundamentally one of poor illumination which can be a function of the survey design.
In this study fault sags within the Hoop Fault Complex of the Norwegian Barents Sea are investigated to gain understanding of why conventional travel-time tomography fails. Through 3D ray-trace modelling it is demonstrated the fault sags can be correlated on illumination maps produced for the originally acquired azimuth. An additional azimuth is modelled, simulating acquisition in the orthogonal direction which demonstrates better illumination in these regions.
Ray attributes allow the generation of synthetic gathers which are time migrated and converted to depth. These demonstrate the sags can be eliminated by the acquisition of an additional azimuth. While the original narrow azimuth 3D survey design is optimal for the shallow targets, if full illumination of the fault complex is a future objective, multi-azimuth or wide azimuth surveys should be considered.
Conventional marine seismic data is affected by the interference from ghosts on both the source and receiver sides. The natural diversity provided by propagation directions, depth variations and imperfect reflections at the sea surface means the notches are not as deep as they often appear after stack. For a flat streamer, the apparent time delay between the main signal and its ghost is angle dependent, and deterministic de-ghosting in the τ-p domain can reduce the effect of ghosts and retrieve the original wavelet spectrum. For a linearly-slanting streamer, further to the angle-dependant time shift a lateral separation occurs in the angle dimension. The amplitude and phase discrepancies around the notch frequencies caused by the variations in depths and effective refection coefficients can be reduced by using a stochastic search for the optimum set of de-ghosting parameters. A deconvolution process stabilized by averaging over a large number of traces in common–slowness panels may be used to address the remaining spectral defects.
Summary We propose a time-domain approach to transform a gather of pre-stack seismic data into an ensemble of highlyresolved traces in the transformed domain. Using a range of various velocity functions in a standard NMO correction routine, we iteratively search for those velocity functions corresponding to the highest ratio of stackable seismic energy among their neighbouring functions, and remove the corresponding energy while updating a stacked trace in the transformed domain. Application of iso-moveout functions helps to avoid the NMO stretching distortions. Application to synthetic and real data shows improvements in resolution and performance. Compared with existing high-resolution Radon techniques, a superior resolution is achieved, resulting in less ambiguous aperture compensation and more accurate reconstruction of stackable seismic events, particularly multiples in the near offset zone.