We Survey Design and Modeling Framework for Towed Multimeasurement Seismic Streamer Data

Transcription

1 We Survey Design and Modeling Framework for Towed Multimeasurement Seismic Streamer Data K. Eggenberger* (Schlumberger), P. Christie (Schlumberger), M. Vassallo (Schlumberger) & D.J. van Manen (Schlumberger) SUMMARY Survey design and modeling is the process of evaluating prior data, if any, so as to optimize the acquisition of a fresh seismic survey. For conventional marine seismic acquisition comprising pressure-only data acquisition, the process of converting geological objectives into a realizable, cost-effective survey is reasonably well understood. However, for towed marine multimeasurement seismic acquisition recording collocating pressure and acceleration measurements, the process is still being explored. This paper proposes a framework of investigation, consisting of three modular workflows to optimize the acquisition geometry and its efficiency, focusing especially on 3D deghosting and wavefield reconstruction algorithms enabled by multimeasurement seismic, without compromising on the survey objectives.

2 Introduction A multimeasurement towed seismic cable as described by Robertsson et al. (2008) acquires pressure from a hydrophone and multiple orthogonally aligned components of particle acceleration. From acceleration,, the gradient of pressure,, can be derived using the relation:, where the dot denotes time derivative. Two axially-orthogonal components are needed to span the space perpendicular to the streamer axis, but only the vertical component, Vz, is required to perform up/down wavefield separation. An equivalent measurement to Vx can be obtained directly from the inline gradient of pressure, P. Perhaps the most obvious application for Vy lies in its potential to double the Nyquist wavenumber in the crossline direction with respect to pressure-only data (Linden 1959; Özbek et al. 2010; Robertsson et al. 2008). This has an immediate benefit in conventional wavefield interpolation but algorithms such as multichannel interpolation by matching pursuit (MIMAP) (Vassallo et al. 2010) are able to reconstruct the total pressure wavefield beyond the Nyquist wavenumber by simultaneously matching both pressure and its crossline gradient, Vy. Joint interpolation and deghosting, whereby the upgoing pressure wavefield is estimated at an arbitrary point within an aperture of multimeasurement streamers, may be achieved through a generalized matching pursuit algorithm (GMP), which simultaneously models the acquired data P, Vz, and Vy (Özbek et al. 2010). Previous work on both synthetic and real seismic data has shown in a qualitative and quantitative manner that these algorithms are robust (Eggenberger et al. 2012; Özbek et al. 2010) also in the presence of noise. However, their performance depends on the streamer spacing, the streamer tow depth, and the frequency content and crossline dip of seismic events, which, in turn, are affected by the subsurface geology. This translates into a dependency on interpolation distance and, therefore, it is desirable to optimize the acquisition geometry before the actual survey, based on survey objectives. Such work is well understood for hydrophone-only acquisition, but less so for acquisition using multimeasurement streamers. In the following, we outline a framework of survey evaluation and design to address this challenge. Such work is based on prior information such as legacy seismic data and models of the physical properties of the subsurface. Those properties then can be utilized in finite difference modeling to mimic the earth response for a range of geometries, to assess the reconstruction performance and to streamline acquisition geometry for the needs of the survey. Framework The proposed survey design and modeling framework for multimeasurement streamer data is shown in Figure 1 and consists of three individual workflows that can be combined and sequenced in a modular fashion. The first workflow analyses legacy seismic data. Its findings then feed into the second workflow that employs seismic ray tracing. Finally, the third and perhaps most powerful workflow deals with seismic forward modeling. The former two procedures were applied for the survey evaluation and design of an experimental towed marine multimeasurement seismic streamer test in the North Sea, the first of its kind, whereas the third workflow was employed on the SEG Advanced Modeling (SEAM) I model using a state-of-the-art finite difference modeling code. Legacy seismic data workflow Often, legacy seismic data are available in survey areas of interest. For the purpose of wavefield reconstruction, these data can be analyzed in two complementary domains: pre-stack and post-stack. The post-stack analysis deals with final migrated 3D volumes of the subsurface where individual regions of the subsurface the reservoir, but also the over- and underburden are investigated in terms of geological features. Figure 2 shows the legacy data analysis performed prior to a towed multimeasurement survey in the North Sea as part of an experimental test. Three horizons of interest, representing laterally changing structures were identified and mapped in 3D because a complex geology can generate out-of-plane energy, detail that is easily lost in a conventional survey. Conventional survey design and evaluation sought to minimise the out-of-plane energy, especially

3 Figure 1 Proposed survey design and modeling framework for multimeasurement streamer data, comprising three workflows that can be combined serially: legacy seismic data analysis, ray tracing, and seismic forward modeling. from the reservoir. This largely dictated the acquisition azimuth, which also compromised the illumination detail. However, survey illumination objectives are not focused solely on the reservoir; it is equally important to have a good understanding of the overburden, as ultimately the drill bit will penetrate those rock formations to access the reservoir. With the recent emergence of 3D deghosting and wavefield reconstruction enabled by towed multimeasurement seismic acquisition, this limitation can be relaxed, allowing more flexible survey design. However, the acquisition geometry must still respect reconstruction quality. Here, the pre-stack analysis of legacy data can give a valuable first insight to the complexity of the recorded wavefield. The inline pressure wavefield is well sampled on most seismic cables, allowing computation of the inline pressure gradient on common shot gathers for input to the MIMAP wavefield reconstruction algorithm. This methodology was first described by Eggenberger et al. (2011). Figure 3 shows a common shot gather and computed inline gradient (top left) with 12.5-m receiver interval. To guide cable spacing selection, both pressure and gradient data were decimated to six input traces of 25-m 125-m sampling, in 25-m increments, and reconstructed using both single- and multichannel matching pursuit algorithms (IMAP and MIMAP). Undecimated reference traces (top right) can be compared with traces reconstructed by IMAP (bottom left) and MIMAP (bottom right). The higher the separation of the six streamers, the more samples will be reconstructed to get back to 12.5-m trace sampling. The uplift from the additional gradient component is evident and may be quantified using repeatability metrics (Eggenberger et al. 2011). The red box highlights the area of interest, corresponding to the three units identified in Figure 2. Ray tracing workflow Figure 2 Inline through legacy dataset. From 3D analysis three areas of interest were identified: near surface heterogeneity, tilted fault blocks around 1-s two-way time and Zechstein salt pillows around 2-s. The legacy seismic data analysis can guide both waveform modeling and ray tracing. The latter can provide a quick look at cable spacing and the bandwidth over which multichannel reconstruction may be expected to provide uplift, as shown in Figure 4, using a 1D velocity function from the legacy data set in Figure 2. The coloured wedges in Figure 4 map the variation of the effective crossline Nyquist frequency, imposed by 75-m cable spacing, in 20-Hz bands from 0-Hz (blue) to 240-Hz (brown). Accelerometer noise increases to lower frequencies, where its use can be limited, so an arbitrary cutoff frequency is shown by the upper white line. A lower white line indicates the threshold where the amplitude ratio Vy/Vz falls below 0.1. The three target units from Figure 2 are indicated as translucent bands edged in green bars. The vertical red line marks the domain of specular reflections in a 1D earth. Crossline dip and diffractions can vary the effective offset so the region beyond the red

4 line is also of interest. More sophisticated 2D or 3D ray tracing can explore a specific structure, if the velocity field is known. However, while ray tracing is suitable for exploring the impact of geometry on aliased bandwidth by analysis of the emerging ray angle, it is not suitable for generating sample wavefields for testing reconstruction. Seismic forward modeling workflow Legacy seismic data can guide waveform modeling as 3D velocity models are often available that approximate the main geological features. Physical property models of the subsurface are also Figure 3. Top left: shot gather of pressure and inline gradient at 12.5-m generated prior to a sampling. Six traces are used to test reconstructions with decimations seismic survey by ranging from 25-m to 125-m. Top right: undecimated reference traces combining seismic, corresponding to reconstruction outputs. (NB data are time shifted that well-log and geological the seabed reflection appears at about 1000-ms.) Bottom left: output information. The third after pressure-only IMAP. Bottom right: output after two-component workflow uses velocity MIMAP. The regions of greatest interest are marked with a red box, from and density models as 1to 3-s and from 25to100-m reconstruction distance, corresponding to input to finite difference the interval from the seabed to the base of unit 3. modeling. Modern computing hardware combined with highly optimized extrapolation kernels makes possible the production-scale use of finite difference modeling. Common applications include the accurate modeling of surveys over complex structures, comparison of different acquisition scenarios for survey analysis and design, and generating synthetic data for evaluation of data conditioning or migration workflows. The modeling software applied employs a staggeredgrid velocity-stress implementation of the Figure 4 Crossline Nyquist frequency as a function of two-way travel time finite-difference timedomain algorithm. It and crossline offset. combines good dispersion characteristics with efficient incorporation of heterogeneous density fields. Because particle velocity is explicitly given by the velocity-stress implementation, it can be sampled efficiently to produce accurate multichannel output (Figure 5). Once shot locations representative to the geological challenge are selected, they are modeled using an initial acquisition geometry and a

5 range of realistic additive noise fields. Ideally, the trial wavefield reconstructions cover single- and multichannel algorithms to evaluate the uplift achieved through the additional acceleration measurements. Reconstruction performance is evaluated using 4D metrics against undecimated reference synthetics in conjunction with spectral analysis and frequency-wavenumber plots in both 2D and 3D. This loop is then repeated on the modeled data for a range of acquisition geometries, changing cable separation to optimise reconstruction performance and operational efficiency while addressing the survey objectives. Figure 5 SEG SEAM I model (left) with one modeled pressure inline (middle) and corresponding Vy zero-offset section for an arbitrary 4C common shot gather. Conclusions Survey evaluation and design for a towed marine multimeasurement survey requires a different approach than for hydrophone-only acquisition, focusing on crossline wavefield reconstruction between streamer positions. To optimise acquisition efficiency without compromising the survey objectives, we described a framework based on three pillars: legacy data analysis, quick-look ray tracing to explore geometry impact, and full wavefield modeling to evaluate reconstruction performance. Production-scale finite difference modeling provides a powerful workflow to optimise wavefield reconstruction parameters and, thus, guide the acquisition effort. References Eggenberger, K., Christie, P. A. F. and Muyzert, E. [2011] Evaluating the benefit of pressure-plusgradient reconstruction of time-lapse seismic wavefields. 73 rd Meeting European Association of Geoscientists & Engineers, Expanded Abstract, H015. Eggenberger, K., Christie, P. A. F., van Manen, D.J., Vassallo, M., Özbek, A., and Zeroug, S. [2012] The fidelity of 3D wavefield reconstruction from a four-component marine streamer and its implications for time-lapse seismic measurements. 82 nd Annual International Meeting, SEG, Expanded Abstract,1-5, doi: /segam Linden, D. A. [1959] A discussion of sampling theorems. Proceedings of the Institute of Radio Engineers, 47, Özbek, A., Vassallo, M., Özdemir, K., van Manen, D.-J. and Eggenberger, K. [2010] Crossline wavefield reconstruction from multicomponent streamer data: Part 2 Joint interpolation and 3D up/down separation by generalized matching pursuit. Geophysics, 75, WB69-WB85. Robertsson, J. O. A., Moore, I., Vassallo, M., Özdemir, K., van Manen, D.-J. and Özbek, A. [2008] On the use of multicomponent streamer recordings for reconstruction of pressure wavefields in the crossline direction. Geophysics, 73(5), A45-A49. Vassallo, M., Özbek, A., Özdemir, K. and Eggenberger, K. [2010] Crossline wavefield reconstruction from multicomponent streamer data, Part 1: Interpolation by matching pursuit using pressure and its crossline gradient. Geophysics, 75, WB53 WB67.