SPECIAL FOCUS: EXPLORATION

The status of wide-azimuth and multi-azimuth seismic acquisition: What is working and what remains to be done

 There’s no doubt about what these methods could do for reservoir imaging and characterization, but can they deliver on their theoretical promises? 

Perry A. Fischer, Editor.

To improve exploration success rates and achieve optimum well placement, better seismic imaging tools are needed. What geologists and geophysicists are learning is that for each geology type, a custom fit of carefully designed data acquisition and processing schemes is necessary for achieving the best answers. One method does not fit all. Rather, geophysicists will continue to evolve a group of tools, each for a specific purpose, including solutions for steeply dipping strata, high-velocity layers (e.g., sub-salt and sub-basalt imaging), highly anisotropic reservoirs and so on. New methods of acquisition that feature wider azimuths and variable azimuths have recently been experimented with, generally, for the purpose of greater reservoir illumination.

The spate of interest in bringing the benefits of wide-azimuth and multi-azimuth seismic acquisition is reflected in the literature of the past few years, especially in the sessions from the most immediate SEG and EAGE Annual Meetings. This article summarizes some of the key findings from those sessions.

INTRODUCTION

Multi-azimuth and wide-azimuth seismic are not new technologies. Land and ocean bottom surveys have been going on for many years. There are now many examples of the benefits that high-fold multi-azimuth and wide-azimuth data can achieve in obtaining a clearer image of the reservoir in these environments.

But with these types of surveys comes a new set of challenges. For example, generally speaking, as azimuths become wider and more varied, anisotropy increases. Also, multiple elimination might become more complex. And the logistical difficulties of coordinating a small armada of source, recording and service vessels, especially in areas with infrastructure and other activity ongoing, can be a daunting task. Despite these challenges, when appropriately applied, these geophysical methods have now been shown capable of improving reservoir imaging and characterization.

Author Rietveld, et al.,¹ said it best at the recent EAGE meeting: “We know from theory and case histories that multi-azimuth data will lead to improved signal-to-noise ratios, improved multiple attenuation and improved illumination. However, because of approximations in current processing technology, the processing of multi-azimuth data will leave errors in the final imaged results, both kinematic and dynamic. Simple stacking of the data, though surprisingly robust in most situations, makes assumptions about data consistency between surveys and will likely not result in the most optimal image.”

WIDE AZIMUTH

In their presentation on wide-azimuth streamer imaging of Mad Dog field,² the authors from BP asked the question, “have we solved the sub-salt imaging problem?” The short answer is “yes.” However, “solved” may be too strong a word; “improved” is a better one. And there are several challenges that remain.

Mad Dog lies in the southern Green Canyon area, in water depths ranging from 4,100 ft to over 6,000 ft as the Sigsbee Escarpment drops off. The reservoir lies under a complex salt body that is close to a rugged water bottom, which results in a velocity field that causes blind spots in the image of the structure. This illumination deficit cannot be filled with conventional surface streamer data.

BP had tried various depth migration routines on conventional towed-streamer data to improve the image, including wave-equation migration and a dual-azimuth imaging method.³ While these had a positive effect, the BP team felt that the reservoir image could be further improved. To do this, the company chose a wide-azimuth design rather than adding additional azimuths. It was felt that since the salt is highly 3D, picking the correct azimuths to add would be a problem.

So in 2004, in what BP called the “first of its kind,” a Wide-Azimuth Towed Streamer (WATS) survey was acquired over the reservoir during a six month period ending in April 2005. The acquisition required two shooting vessels on the side of one cable vessel (which could have a source on it as well, but didn’t), with one of the source vessels at the front of the cables and one at the back, Fig. 1. Sources were sequentially fired in a repeating pattern. Each line was shot four times, and the streamer vessel was offset by a kilometer each time. The cable vessel had 8 x 8,100-m streamers separated by 125 m, so the synthesized shot patch was 8.1 km by 4 km. This would achieve a broad range of azimuths and offsets. For other reservoirs, the layout is easily modified by moving the shooting vessels relative to the cables to suit the imaging requirements.

Fig. 1. Wide azimuth (left) vs. narrow azimuth, as used at Mad Dog.

The survey was designed using BP’s finite difference modeling software⁴ to obtain the optimal image quality at the lowest possible cost. Another acquisition design that also might have worked was an array of autonomous ocean bottom seismic recording devices, accompanied with a dense, wide shooting pattern. These were later built and 900 were deployed at BP’s Atlantis field in late 2005 to early 2006. Results at Atlantis are good.⁵

At Mad Dog, to process the data, conventional approaches worked most of the time, but there were several unique considerations:

• Water column statics are a concern. A primary difference between WATS and a conventional narrow-azimuth survey is that the conventional survey has single-fold crossline coverage, while the WATS survey has higher crossline fold. This requires taking “a more surface consistent approach rather than the more typical base trace methods typically employed.”²
• For this type of data, tomography needs to be completely re-thought, say the authors. With narrow-azimuth surveys, azimuth can be ignored or simplified by various data regularization methods. With wide-azimuth data, the crossline offset information is used to build a better velocity model. At Mad Dog, BP used something called vector offsets. Vector offsets have both inline and crossline components. Applying the method of Ledart,6 BP calculated vector-based residual move-outs and maintained the wide-azimuth character of the data for the sediment velocity inversion. Given the sediment velocities, salt model building continues as usual, with an iterative top-down approach.
• Migration of WATS data is similar to conventional survey data except for the size of the data set. For example, conventional narrow-azimuth data could have 40 offsets migrated using Kirchhoff prestack depth migration, but the WATS data might have as many as 320 offsets, and this would be costly to run. Alternatives to Kirchhoff are wave-equation imaging methods, including techniques such as shot migration, plane-wave migration and delay-line migration.

Multiple elimination was challenging at Mad Dog because of the large shot increments, for which conventional SRME is less appropriate. Multiple diffraction can become significant in areas where residual-multiple energy left after processing is of similar or higher amplitude to primary reflections.

However, multiple attenuation can be one of the major advantages of the wide-azimuth geometry, but shooting density matters. Modeling by Ceragioli, et al.,⁷ found that there was considerable benefit in what they called a “truly” three-dimensional stack. “In a complex geological setting, the kinematics of multiple events vary along different azimuth directions. Free-surface multiples, even multiple diffractions, are therefore attenuated while the primaries stack constructively during a pre-stack migration process.” To better understand and quantify this, the authors built a simple model made of a dipping reflector and a diffracting point, and they studied the surface multiple between the reflector and the diffractor. “Results show that the stack process alone can provide for an additional attenuation of 15 dB when stacking dense 3D WATS data compared to a classical 2D stack.”

Further, Ceragioli et al. concluded that “when the application of 3D SRME is envisaged in the processing sequence, extreme care must be given in the choice of the crossline sampling, as the correct application of 3D SRME may require a dense sampling in the crossline direction. This requirement can be the major constraint in the definition of acquisition parameters for a wide azimuth marine survey.”

Xia and Matson⁸ concluded that “the WATS data with rich diversity of azimuth and offset is effective at attenuating multiples-through simple stacking of post-migration image gathers. However the effectiveness of stacking degrades for exploration scale sampling. Moreover, simple stacking precludes prestack analysis for velocity estimation and angle dependent amplitudes on the image gathers. Therefore, it is desirable to develop better techniques to eliminate multiples.” They adopted the extended wavefield extrapolation method that was developed initially for OBS data to accommodate the WATS geometry. See their paper for further details. Mad Dog, of course, isn’t in the exploration mode, but the point is still valid.

All told, the changes that were implemented at Mad Dog, whether surface statics, tomography, multiple elimination, or others, were successful. The team concluded that “preliminary images using the old velocity model derived from the dual-azimuth imaging project indicates that we have achieved a step change in image quality at Mad Dog. By acquiring the data in a manner that meets the geophysical needs of the problem, we have greatly enhanced the image of the reservoir.”²

MULTI-AZIMUTH

Like wide azimuth, the need to acquire a multi-azimuth marine survey must be compelling in order to justify the added cost. However, poorly imaged areas of a given reservoir may have no other way to be seen. In their presentation, “Key aspects of multi-azimuth acquisition and processing,”⁹ authors Keggin et al. discuss what they have learned about the subject. Modeling indicates that there are benefits to geometrically symmetrical azimuth design. An ideal multi-azimuth (MAZ) marine acquisition will have azimuths that are regularly sampled from two to six directions. The authors emphasize two benefits of regularly sampled azimuths: The potential to measure and correct for azimuthal anisotropy, and multiple diffraction attenuation.

Multiples can become a problem in these situations, because the acoustic energy reflected from the reservoir is typically poor, which is what prompted these more difficult acquisitions in the first place. Thus, the energy from multiples can easily be as strong as, or stronger than, the primary reflections. A method from Keggin, et al.,¹⁰ suggests that primary reflection data will tend to add constructively and the diffraction multiples will cancel. This can be achieved with towed streamer MAZ acquisition as well as with seafloor nodes.

MAZ acquisition. For conventional towed streamer acquisition, each multi-azimuth survey is acquired with a single vessel. The vessel then turns and repeats the acquisition over the target in several directions. The regular variation in source-receiver azimuth recorded this way can be exploited to form true azimuth data sets. The BP and PGS authors state that “when acquiring these surveys, it has become common practice to include azimuth as a binning criterion. For conventional wide-tow marine seismic, a range of source-receiver azimuths is acquired, more so for the near traces of the far-streamers.

“Consider a typical 10 streamer layout with 100 m streamer separation and 250-m inline offset to the near-trace, [Fig. 2]. For this layout, it can be seen that the majority of data is recorded within +/– 15° azimuth of the sail-line direction. The yellow highlights indicate that portion of the recording spread that has larger azimuths. In this example, the azimuth of the near-trace to far streamer is about 60°. This reduces to 15° at approximately 1,700 m on the far cable. These traces are used in conventional single azimuth surveys but are dropped (yellow highlight) during multi-azimuth processing leaving coverage holes. However, these holes can be filled by the same traces from different recording azimuths. This illustrates how a survey at 60° to the first can contribute its port-cable near-traces to the azimuth of the first survey.

Fig. 2. Wide-tow azimuth range.9

“[Figure 3] shows the near-offset coverage for a three-azimuth survey. The thick lines represent the east-west (left-to-right) survey with azimuths greater than 15° rejected. The thinner blue lines represent contributions to that source-receiver azimuth acquired by sail-line azimuths of +/– 60° to this. The three sail-line azimuths used for this survey were zero, + 60° and – 60° (counting the east-west survey as zero azimuth). This “ball of string” coverage raises the possibility of relaxed infill criteria, providing that the relevant data can be viewed on board for QC during acquisition.” Relaxed infill requirements could translate into lower-than-expected acquisition costs. In addition, the procedure of selecting true source-receiver azimuth ranges presents perhaps an unconventional result for processing to regularize prior to artifact-free 3D migration.

Fig. 3. “Ball of string” coverage for near-offset range.9

MAZ processing. Specific processing techniques developed for multi-azimuth data sets include multi-survey binning, azimuthal regularization, velocity and stacking. There is a question as to whether there is benefit in sorting into source-receiver azimuth data sets and whether there are any advantages other than the potential to measure azimuthal anisotropy.

The possibility to quantify azimuthal anisotropy raises much interest in the geophysical community. For some years it has been stated that azimuthal anisotropy is likely to be present in this data.¹¹ Most multi-azimuth data sets have been processed using pre-stack depth migration, but could time-migration be suitable for these data sets? Processing experience shows that, indeed, pre-stack time migration can be successful in producing a good multi-azimuth image and give a significant uplift over single-azimuth data.

Data summation is another issue. While noisy areas could benefit from a straight stack, what if one azimuth in particular illuminates the target best at a particular location? The summation of all azimuths is likely to produce an inferior result. The question of how many azimuths are required is essential and one is best answered prior to acquisition. A minimum of perhaps three would be required to pick azimuthal anisotropy, but what improvement can be expected by adding more? Answers to these questions may be quantified by use of specific metrics, such as signal-to-noise, coherency, multiple content, etc.

The issues of optimum stack and signal-to-noise ratio are not new and perhaps require further work to be sound enough for production processing. However, practice has shown that more simple processes can give good results. A good proportion of multi-azimuth data sets processed to date, both time and depth migrated, have benefited from ‘trim statics’ prior to summing each azimuth. Experience shows that summing without such a process can be surprisingly robust, and that trim statics can give a further uplift if suitably controlled. Indeed, this first-order approach has formed an integral part in multi-azimuth processing to date, both as a final touch and as a key step in the process of achieving an optimum stack.