2012 SEG SEG Las Vegas 2012 Annual Meeting Page 1

Transcription

1 Survey design for high-density, high-productivity, broadband vibroseis point-source and receiver acquisition a case study Peter van Baaren, Roger May, Alexander Z,arkhidze, David Morrison, and John Quigley, WesternGeco; Abdulla Al Qadi, Crescent Petroleum, Sharjah Summary In this paper, we describe an integrated design of acquisition parameters for a 3D seismic survey in a difficult geological and logistical environment using an ultra-highchannel-count point-receiver recording system, productivity enhancement techniques, and advanced processing tools. Advances in acquisition hardware and techniques change the way seismic surveys are designed. Such advances include continuous recording, faster computers that allow running more elaborate algorithms, and new techniques such as surface wave methods to model and subtract aliased ground roll. These new tools required different input data and different sampling. How these advances influence the design process is illustrated using a case study for a 3D survey acquired in the 3 rd quarter of 2011 located in a thrust belt area. This exploration play extends from the United Arab Emirates to Oman where the prolific stable Arabian plate was deformed by the Late Cretaceous orogeny, resulting in structurally controlled traps. Typically, the imaging task in thrust belt regions is challenging due to the complex raypaths and steeply dipping events. In addition, much of the survey areas is covered by an inhomogeneous sand layer. The seismic data were acquired using a high-productivity technique that uses time and distance rules coupled with integrated QC rules based on active spread criteria to maximize productivity and minimize crosstalk between the individual seismic sources. The data were acquired with single 80,000-lbf vibrators using a special sweep to generate useful energy below 6 Hz. The initial data from the 3D survey shows that the proposed integrated 3D survey design is adequate for coherent noise attenuation, source crosstalk removal, and building a nearsurface model. Introduction The advent of ultra-high-channel-count systems makes the use of point receivers economically feasible and allows for adequate sampling of both signal and noise. Point-receiver acquisition opens up many new opportunities for processing the data, ranging from productivity enhancement techniques, new methods for characterizing the near surface and coherent noise separation techniques, to complex imaging. The fundamental criteria are not changed; the survey design process must ensure that signal and noise are sampled well enough to fulfill the requirements of both the noise attenuation and signal processing algorithms. The design must also consider operational issues and cost. The survey size must be as small as possible, but large enough to image the targets. Once these key parameters are determined, others must be considered. These additional, different requirements will be discussed in this paper. A rich 2D point-receiver data set was used as input into the 3D design process. These 2D data were acquired based on the results of the D seismic campaign in the adjacent Emirate of Dubai (van Baaren and van Kleef, 2008) using point-receiver acquisition. These 2D data are inherently limited in their ability to properly position events that have an out-of-plane dip, thus, a 3D survey was required to further explore the area. The misties with the 2D data are shown in Figure 1. Figure 1: Ties between group-formed point-receiver 2D lines south of the 3D prospect area. Note the large misties. Survey size One of the key parameters to determine in the survey design stage is the size of the survey as this has a large bearing on the cost. A sufficiently large area must be covered to ensure that all reflection energy from the targets is captured and migrated to the correct location within the constraints placed by the geology and land access. The Crescent Petroleum concession is wedged between the Emirate of Dubai and the outcrops of the Hajar Mountains. The area has a proven active petroleum system, so the exploration task is to find potential traps. A number of wells have been drilled in the concession, starting in the mid 1960s, but have failed to find commercial accumulations. These were all based on 2D data. It is becoming common practice to perform advanced forward modeling in complex environments to determine the optimum survey size. However, in the exploration stage, it is somewhat a Catch 22 situation. To define the model requires a good understanding of the structure, but, if we knew the structure well enough to model it, then we would not be at the exploration stage. In the absence of a realistic model, two methods were used to determine survey size. The first method was to use a 3D dip map produced by interpolating the sparse grid of 2D data. This was in the form of x- and y-coordinates, target depth, and dip values throughout the 2D area. The target zone was delimited by using hydrocarbon down points, or SEG Las Vegas 2012 Annual Meeting Page 1

2 spill points, giving a contour of the target area. Several leads (roll-over structures related to the thrusting) were identified at different depths. For each grid point that fell in these leads, the image point was projected to the surface using a straight ray approximation. The survey boundary would be the polygon that captures these points plus a suitable fold halo. Another way to determine the survey size is to look at the size required for the migration operator to collapse diffractions and/or Fresnel zones. Contours of the leads were provided by the interpreter derived from 2D and gravity data. The closure of the structure in the west is determined by a thrust fault and the position of the fault must be known accurately, as this would determine if there was a trap. In the east, fewer dips were expected, and the size would be determined by the depth of the oil/water contact. So, based on these contours produced using the method described above, a 30 aperture based on the size of the migration operator (Cordsen et. al., 2000) was used on the west and a compromise solution of a 10 aperture, based on the size of the Fresnel zone and migration operator size in the east, was used to determine the desired image area. It was decided to overlap the fold taper and migration aperture, ensuring that the image area would be full fold for 4 km offsets and smaller to ensure as large an image area as possible. All projected image points fall within the survey area defined in this way giving confidence that the survey area was correctly sized. Other criteria were also considered. For example, to allow the tomographic velocity model building to converge, the survey had to have a minimum width to allow the lowfrequency component to be correctly solved. At 13 km at the widest point, the survey was wide enough to meet this criterion. Geometry The advent of ultra-high channel count continuous recording systems and new coherent noise removal techniques allow great flexibility in designing full-azimuth survey geometries. The aim is optimal noise attenuation and dense signal sampling with a minimal footprint. The footprint is caused by variations in offset and azimuth sampling, and hence, different filter responses of the many processing steps that have a dependency on these parameters. The footprint has been described amongst others by Bianchi et al. (2009). The optimal solution would be a dense carpet of shots and receivers. As the benefits of full-azimuth acquisition are well documented (Vermeer, 2002), the complex structure with dips in all directions made no explanation necessary for the decision to record a square patch and symmetric sampling. The key parameters left to decide are the offset and the source and receiver line density. As the survey area is covered by a thick blanket of sand, lines must be cleared for access, thereby limiting the survey designs to discrete source and receiver lines. These discrete lines allows easy gathering of traces in cross spreads. Cross spreads allow optimal noise attenuation where ground roll is smoothly varying from shot to shot. In addition, for point receivers, well-sampled cross spreads are required for ambient noise suppression, as transient noise will appear as burst noise on single traces in the common-receiver direction. The spacing between the source and receiver lines was 200 m x 200 m, which is based on a desired trace density and overall project goals and time frame. The target depth is in the order of 4 km, and using rules of thumb, the maximum useful offset should also be in the order of 4 km (as minimum). However, this does not take into account the complex raypaths. The study of parameters of the 2D data acquired in the past showed that longer offsets produced more coherent results at the zone of interest. Decimation tests on 2D data showed that offsets up to 6.5 km still appear to contribute to improving the reflection image. Beyond this, no noticeable difference was seen. Hence, it was decide to use 6600 m as a maximum inline and crossline limit as this is a multiple of 200 m (the chosen source and receiver spacing). Spatial sampling A number of criteria exist for spatial sampling. Obviously, reflections from the steeply dipping events and the diffractions tails coming from complex faults must be properly sampled to be collapsed in migration. In addition, aliasing of the migration operator should be considered. The sampling locations of the point-source and receiver measurements were derived from the 2D point-receiver measurements. This was 20 m (post digital group forming), which was previously shown to be adequate sampling for the steeply dipping events and the migration operator. Based on this, a sampling of 12.5 m inline and crossline was chosen as the coarsest sampling. With point-receiver acquisition, the requirement is to sample the noise adequately for the chosen noise attenuation technique. The 12.5-m sampling was adequate both inline and crossline, as all of the modes of ground roll are sampled either unaliased or only mildly aliased. Mildly aliased means the noise does not wrap back into the signal space and can be removed without damaging signal using the latest methods for ground-roll attenuation. To avoid introducing azimuthally variant amplitude and phase distortions using either digital or analogue arrays, the pre-migration processing flow must be designed to work with an inherently lower signal-tonoise ratio of the point-receiver/point-source traces. This requires different or at least modified processing techniques. For example, one may use deconvolution techniques that can handle poor signal-to-noise ratio inherent in point-receiver/point-source recording (Kirchheimer and Ferber, 2001). Additional noise attenuation might be achieved by the migration process. Having 12.5 m sampling in both source and receiver domains and 200 m line spacing provides a very high trace density, as well as full offset and azimuth sampling, as input into the stack and the prestack migration processes. The high trace density is effective as reflections not visible SEG Las Vegas 2012 Annual Meeting Page 2

3 on the raw records appear in the stacked data and this should be further improved by migration. Near surface It was expected that the near surface would be very inhomogeneous with short-period lateral velocity and elevations changes, and that this would have a large effect on the final image after migration. One method to determine the near-surface properties is to use shear-wave inversion methods that use ground roll to build near-surface velocity models (Strobbia et al., 2010). The best data to run this process are point-receiver data that measure the ground roll with high fidelity and low frequencies free of amplitude and phase distortions caused by receiver arrays. Low frequencies are required to obtain information from reflections as deep as possible. Hence, low-frequency sweeps and point receivers were required to build the nearsurface model. The near-surface velocity model and the data after application of the statics derived using this method are shown in Figure 2. Figure 2: Near-surface velocity (P-wave) profile derived from ground-roll measurements (top) and near-surface section with these static derived (below). How many channels? The aim of the survey design process is to determine a costeffective solution that meets the technical requirements. This requires analysis of the key cost drivers. One option was to deploy 80,000 channels to span the full block width with enough additional channels for effective shooting, and this was possible with our system. However, this may not be the most cost-effective solution as total acquisition time and crew costs are key variables in the cost equation. Acquisition time is determined by the rate of progress of the slowest component in the operation, not just how fast the vibrators can go. The crew cost is proportional to how much equipment and how many people are involved. The speed the survey can be acquired is limited by a number of factors such line clearance, spread roll, shot cycle time, and operational hours. The use of high-productivity techniques with an adequate number of vibrators can reduce the source cycle time to less than the sweep length and listen time, but this is only of benefit if other aspects of the operation can keep up. For this project, there were limitations on how many bulldozers could realistically be mobilized and drivers trained. There were also limitations on how many people could be hired, trained, and managed. Safety is paramount and the cycle of sourcing, evaluating, training, supervising, and managing new resources must be respected. Having established reasonable parameters for what could be achieved in line clearance and line layout, different acquisition techniques and methods were modeled to determine the most cost-effective solution. As a result of this modeling, a design utilizing a source repeat factor of 2 was selected, which required only 40,000 channels for efficient recording. This doubles the amount of VPs required, but the length of the job was being determined by the time required to clear the lines and lay out the channels. Source productivity enhancement techniques enabled shooting the required increased number of VPs within the planned time period. Productivity enhancement A number of different techniques such as slip-sweep, independent simultaneous sources and distance separated simultaneous sources (DSSS), are used currently and are well documented (Rozemond, 1996; Bouska, 2008; Howe et al., 2008). The choice of technique depends on several factors including block width, how efficiently the vibrators can move around, and how many obstacles and how they are distributed. For example, with a block width of only 13 km and long offsets required, the use of DSSS was not an option. A key concern of all these methods is crosstalk from other vibrators and/or harmonic effects. This must be balanced against the design that required a square patch, fine sampling, and long offsets. The survey would not be economically viable without the use of enhanced source productivity techniques. The modeling referred to above was utilized to determine the most cost-effective solution. As a result, this survey was shot using the managed spread and source technique. This technique is based around optimizing the sequence of acquisition of vibrators that are in position and ready to record, with certain rules. These rules could be an offset rule, (vibrators must separated by a certain distance), time rules (there must be a certain minimum period), spread criteria rules, or a combination of rules. In the sequence, only sources that have the spread available and meet all defined criteria are considered and prioritized. These rules and implementation are managed automatically by the recorder with no operator intervention. Using the start times from each of the sources, the data are extracted from the continuous recording operation. The rules used were 6 s slip time (equal to the listening time) when offsets were shorter than 6.6 km and 3 s if the offset was greater than 6.6 km. The sweep used was an 18 s maximum displacement sweep (Bagaini, 2008). The 6 s slip time was chosen so that, within one listening time and within the maximum desired offset, there would be no overlap of fundamental energy. For offsets greater than 6.6 km, the slip time was chosen to be half the record length to put high-energy fundamental noise at least half way down the record where there would be a significant difference in amplitude level between the desired energy and crosstalk. As the vibrators were distributed across the block both north and south of the spread, and each vibrator SEG Las Vegas 2012 Annual Meeting Page 3

4 moves up at its own speed, the sequence that the vibrators were shooting was a random pattern. This random sequence ensures that interference crosstalk is random and of different energy levels in a common receiver direction, and hence, can be readily attenuated in processing. The same also applies to harmonic energy. An example of data with interference noise and after attenuation of this noise is shown in Figure 3. The chosen design results in strong coherent events already visible at stack level, demonstrating that the chosen spatial sampling intervals are adequate for the specific structural imaging objective. The chosen parameters allow generation of a spatially coherent near-surface map that appears to follow surface features and outcrops. The chosen high-productivity technique allows for economic acquisition of full-azimuth dense geometries on land whilst facilitating effective removal of crosstalk between sources noise. Due to the structural complexity, full evaluation will be required after earth model building and depth imaging. Acknowledgements Figure 3: Figure showing a far cable (6 s of data) from a common shot gather after spherical divergence correction with interference from other vibrators before simultaneous interference removal (left) and after (right) Results The survey was acquired as designed and the initial results show much promise in the field stack generated during the acquisition. The preliminary stack is shown in Figure 4. From this initial stack, it is clear that there are coherent events with conflicting dips in the shallow section and deeper layers. These events are not visible on individual shot records. With such complex structure, determination of the earth model is a critical step in the success of the survey. Additionally, application of the latest anisotropic reverse-time migration techniques is a requirement. This work is currently ongoing. The parameters for the survey are shown in Table 1. Based on the data acquired and analyzed, the following conclusions can be made. The authors thank the management of Crescent Petroleum and WesternGeco for permission to publish this paper. The authors also acknowledge the many people who contributed to this project, from reviewing, to asking critical questions, and providing advice. Table 1: Acquisition parameters. No. of channels per line 1056 No. of active lines 33 Spread type 1 line roll symmetric split spread 66 line emulation Nominal channels live 34848* Receiver interval 12.5 m Receiver line interval 200 m Source interval 12.5 m Source line interval 200 m Nominal fold 6.25-m x m bins Source repeat factor 2 Roll factor VP per 2 receiver (*) Up.to 40,000 channels were live during acquisition to account for the superspread required to shoot the survey effectively Figure 4: Field stack produced (TWT) by the field crew showing complex structure binned at 12.5 m x 12.5 m without any digital group forming. SEG Las Vegas 2012 Annual Meeting Page 4

5 EDITED REFERENCES Note: This reference list is a copy-edited version of the reference list submitted by the author. Reference lists for the 2012 SEG Technical Program Expanded Abstracts have been copy edited so t hat 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 Bagaini, C., 2008, Low -frequency vibroseis data with maximum displacement sweeps: The Leading Edge, 27, Bianchi, T., D. Monk, and J. Meunier, 2009, Fold or force: 71st Conference and Exhibition, EAGE, Extended Abstracts. Bouska, J., 2008, Distance separated simultaneous sweeping: The world s fastest vibroseis technique: Vibroseis Workshop, EAGE, Expanded Abstracts. Cordsen, A., M. Galbraith, and J. Pierce, 2000, Planning land 3 -D seismic surveys: SEG Geophysical Developments Series No. 9. Howe, D., A. J. Allen, M. S. Foster, I. J. Jack, and B. Taylor, 2008, Independent simultaneous sweeping: 70th Conference and Exhibition, EAGE, Extended Abstracts, B007. Kirchheimer, F., and R. Ferber, 2001, Robust surface-consistent deconvolution with noise suppression: 71st Annual International Meeting, SEG, Expanded Abstracts, Rozemond, H. J., 1996, Slip-sweep acquisition: 66th Annual International Meeting, SEG, Expanded Abstracts, Strobbia, C., P. Vermeer, A. Laake, A. Glushchenko, and S. Re, 2010, Surface waves: Processing, inversion and removal: First Break, 28, No. 8, van Baaren, P., and F. van Kleef, 2008, Single -sensor vibroseis acquisition in complex thrust belt areas: A case study from Dubai: 78th Annual International Meeting, SEG, Expanded Abstracts, Vermeer, G., 2002, 3D seismic survey design: SEG. SEG Las Vegas 2012 Annual Meeting Page 5