Bathymetry map of Gulf of Mexico showing location of Lower Tertiary Jack and St. Malo fields in Walker Ridge. Also shown is Mad Dog in Green Canyon, where the first wide-azimuth survey was acquired in the Gulf of Mexico. (Images courtesy of Chevron)

Since its discovery in 2004 in the subsalt Walker Ridge Protraction Area, the Jack field has come to represent the emerging Lower Tertiary Trend in the deepwater Gulf of Mexico (Figure 1). The large salt-cored four-way structure made news in September 2006 after successfully completing a record-setting production test on the #2 well in around 7,000 ft (2,135 m) of water and at reservoir depths of around 27,000 ft (8,235 m). During the extended test, which was designed to evaluate a portion of the total pay interval, the well sustained a flow rate of more than 6,000 bo/d, confirming the flow capability of this frontier area.

As the challenges of subsalt exploration fade, answers to new and increasingly detailed questions are sought in an effort to minimize risk and optimize project value at Jack. However, with only two penetrations, the field possesses many unknowns whose descriptions are severely constrained by the available seismic data quality. Key uncertainties such as faulting, compartmentalization and overall structural geometry all play a prominent role in determining well placement and well count, which are factors that impact ultimate recovery, development scenarios and project economics. Our basic assertion is that in a deepwater sparse-well setting, better seismic leads to better decisions. Hence, we should do everything possible to improve the quality of seismic data.

The seismic data quality covering the field is typical of many subsalt images. Data at the reservoir level are low frequency (~10 Hz dominant) and contaminated with remnant noise, making characterization of key reservoir uncertainties, e.g., faulting, an onerous undertaking at best. Over much of the Jack structure, the water bottom and top salt are separated by ~1,500 ft (457 m) of sediment, while the reservoir is beneath an undulating salt layer that can exceed 10,000 ft (3,050 m) thick. The frustrating result of this configuration is that the family of water bottom and top salt multiples coincide with and contaminate primary energy at the reservoir level. Efforts to attenuate this noise can be at the expense of the primaries or result in remnant multiples that still overwhelm the signal. Needless to say, given few wells and a feeble image, characterization of the Jack reservoir presents a considerable challenge.

Recently, the industry has recognized a step-change improvement in subsurface imaging in deepwater subsalt settings with wide-azimuth (WAZ) towed streamer acquisition and processing. Presented with the opportunity for a similar improvement in the data quality over the Jack field, an integrated modeling effort was undertaken which included simple 1-D synthetics, 3-D ray tracing and 2.5-D finite difference models. While these analyses facilitate the understanding of the advantages as well as the limitations of proposed WAZ survey designs, they are not able to directly aid in the determination of the value that the improved data quality may have to the project. To help in answering the value of information (VOI), a parallel reservoir modeling project was conducted to address the implications of acquiring a WAZ survey prior to development decisions. Together these models provide an integrated analysis that can be used to effectively communicate the impact and value of a WAZ investment early in the appraisal phase of a project.

Modeling to understand WAZ at Jack

Details vary, but the general WAZ survey design involves source boats positioned some distance laterally from the streamers to acquire crossline as well as inline offset. One pass with this configuration comprises receiver coverage with a particular crossline:inline offset aspect ratio, sometimes referred to as a “tile.” The entire acquisition is then rolled, or repeated, by moving out the relative distance of the sources from the receivers to increase the crossline:inline aspect ratio. Compared to a narrow-azimuth towed streamer (NAZ) acquisition, a WAZ survey is shot over a longer duration of time but can achieve considerably higher trace density. The associated non-trivial increase in cost necessitates a thorough understanding of the cost-benefit trade-offs among WAZ designs relative to more conventional NAZ designs.

Illumination and 1-D Models. Ray tracing through an earth model in various acquisition directions can establish the most advantageous shooting direction and ascertain the influence of complexities in the overburden, such as salt, on the ability to image the subsurface. Still, ray tracing for illumination provides only a measure of quality based on the ability of a ray to travel from source to receiver through a given earth model, and it does not lend any insight into how noise (e.g., multiples) interacts with the signal. A 1-D model provides a relatively simple and quick method to examine the interplay of primaries and multiples among different survey designs at the Jack field, from which we concluded that stacking traces that were acquired with a increased crossline:inline offset ratio is an effective form of demultiple in itself. Many complex and exhaustive modeling techniques exist to fully assess the spectrum of WAZ acquisition designs. However, timelines and other business drivers dictate that practical compromises are exercised while maintaining the scope of the project. In this paper we examine the results of the 2.5-D finite difference models.

Finite Difference Models. Finite difference modeling hones in on specific parameters such as sail line spacing and crossline offset. Results corroborate the 1-D model findings: Increasing the crossline offset decreases the coherent noise in the raw stack. Figure 2 shows images of the finite difference models with NAZ and WAZ geometries. A noteworthy observation is the reduction in “wavefronted” multiples emanating from the bright subsalt horizon on the NAZ versus the WAZ (Figures 2a-b). The NAZ pattern of these multiples overlain on the primaries may resemble the geometry of a faulted anticline, which is a reasonable geologic scenario, if not a potentially misleading interpretation. As the crossline:inline aspect ratio approaches 1 — in other words, a more truly 3-D acquisition — signal-enhancing processes achieve superior results. The WAZ acquisition rectangle, as opposed to the NAZ acquisition line, also reduces the approximation and interpolation requirements for 3-D surface-related multiple elimination (SRME). Perhaps most remarkably, stacking across an acquisition rectangle rather than a line becomes a highly effective form of demultiple in itself.

Figures 2e and 2f show migrations of synthetic shots generated without free-surface multiples (interbeds still present). Even though they respectively represent NAZ and WAZ geometries, image quality is virtually equivalent, testifying to the profound negative impact of free-surface multiples on subsalt images. In environments such as Jack, where poor illumination is not the critical imaging issue, addressing the multiples that interfere with the subsalt signal introduces a huge advantage. WAZ provides a mechanism to reduce the influence of this potentially detrimental source of noise from the start.

Value of information

What is the value of a WAZ survey at the pre-development phase of the field? We approach this question from a different angle. Rather than asking what the value is by invoking an endless array of what-if scenarios, we ask what it needs to be in to order optimize a particular development parameter, in this case well placement. An oft-cited statement is if better seismic data leads to an interpretation that can prevent the drilling of a dry hole, then the survey adds value. We examine the more subtle outcome of an uneconomic well drilled near a baffling fault unseen on NAZ seismic.

In the current reservoir model a P10/P50/P90 range of fault patterns exists based on different interpretation realizations. Given a grid of wells, there is a probability that a few will be drilled near a fault that might negatively impact total production. We simulate the scenarios of producing an “average” well drilled near a fault as compared to that well drilled away from a fault. Both wells come on line at reasonable rates, but over time the rate of the well near a fault plummets as the fault restricts oil production (Figure 3a). This negatively affects recoverable reserves and, in turn, net present value. A WAZ survey will likely not delineate every fault perfectly, but so-called “perfect information” is not required to recoup the cost of the survey. If the placement of fewer than 20% of the wells affected by faults could be optimized based on the
WAZ image, the survey realizes its value in time for development decisions (Figure 3b). As the overall image quality is improved, WAZ value may be realized elsewhere throughout the life of the project, e.g., improved definition of the salt model, identification
of drilling hazards and delineation of zones within the reservoir.

Conclusion

Optimal WAZ survey design and accurate prediction of image quality are critically reliant on the use of routine finite difference modeling on a case-by-case basis. Ray tracing, 1-D and 2.5-D finite difference models confirm the potential increase in image quality that a WAZ survey can yield at the subsalt Jack asset.

Industry’s progression toward development of subsalt fields demands a progression of technology to meet the challenge. A WAZ survey can contribute to a higher quality subsalt image and a subsequently clearer understanding of structure, faulting and perhaps stratigraphy. Increased confidence in interpretation can lead to improved characterization of risk and uncertainty, the effects of which propagate through the entire appraisal process and impact structure, static and dynamic models, and ultimately project economics.

Acknowledgements

We thank Chevron North America E&P for permission to publish this paper. We also thank the Chevron Jack team and our partners, Statoil and Devon. Bruce VerWest at CGGVeritas provided invaluable discussions on the modeling process and results.

Editor’s note
This article is an excerpt from a longer article that ran in the November 2007 issue of The Leading Edge and has been reprinted with permission from the authors.