ONBOARD SEISMIC PROCESSING SPEEDS EXPLORATION PROGRAM OFF INDONESIA

D. Gordon Armstrong
Atlantic Richfield Indonesia Inc.
Jakarta

Donald F. Dean
ARCO Exploration & Production Technology
Plano, Tex.

Jeffrey C. Porter
Landmark/Asia Pacific
Singapore

Jay E. Valusek
Landmark Graphics Corp.
Houston

During acquisition of a large 2D seismic survey in the Java Sea this spring, Atlantic Richfield Indonesia Inc. and Landmark Graphics Corp. successfully completed the industry's first known full-scale onboard processing project.

Using powerful interactive workstations, a team of geophysicists fully processed 4,000 km of data before the ship docked at the end of the 7 week survey. To our knowledge, nothing comparable has been done before, in terms of the quality and quantity of data processed onboard an acquisition vessel.

Since the advent of digital processing technology in the 1950s, seismic acquisition, processing, and interpretation have been sequential, largely isolated steps in the exploration process. Full onboard seismic processing allows them to take place almost concurrently-with significant results.

Not only did onboard processing trim a huge amount of time off ARCO Indonesia's exploration program, but it allowed ARCO's interpreters to modify the acquisition design to obtain the seismic data they needed, without having to shoot additional lines in the future.

In this article, we will explain why and how we achieved this milestone, and what some of the implications may be for ARCO and the industry.

OBJECTIVES

In mid-1991 Pertamina, the state oil company of Indonesia, approached ARCO Indonesia to look for ways to speed up development of natural gas reserves in the Java Sea, one of the country's hottest exploration areas.

Oil consumption at home and increasing demand for electricity on the densely populated island of Java have begun to threaten Indonesia's future as an oil exporter. According to a recent study, Indonesia could become a net importer of oil by 2000 if it does not maintain enough new discoveries to match reserve depletion.

Due to an exploration slump from 1986 to 1989, only two major new field development projects are scheduled to begin production in 1992-93. By bringing more natural gas on stream as quickly as possible, Pertamina hopes to meet the nation's growing need for electricity while preserving more of its crude oil for export.

Beginning in 1992, ARCO's geoscientists were planning to drill many new gas exploration wells in the Ardjuna basin of the northwest Java Sea, based on several vintages of old, fairly low quality seismic data. But they wanted to acquire, process, and interpret a comprehensive new regional 2D survey in time to influence as many of those well locations as possible (Fig. 1).

Unfortunately, the new survey could not begin until spring of this year, and acquisition alone was expected to take until the end of June. In addition, because marine acquisition in the Asia-Pacific is at record levels, processing turnaround from Indonesia's overloaded geophysical contractors can take 8 months or more for a survey of this size.

Under these circumstances, the new data would not have been ready for ARCO's interpreters until the spring of 1993. And only after completing a new interpretation would they be able to make better-informed drilling recommendations. That's why last fall, ARCO began exploring the feasibility of onboard seismic processing to cut the time between acquisition and the drilling decision. The goal was not to improve quality control of field acquisition but to process the survey onboard the ship.

Typically, acquisition quality control systems "process" only a small subset of the total volume-one or two lines, or every 10th line-to monitor data quality in the field. ARCO, in fact, uses very strict control measures to ensure acquisition of the highest quality data possible.

But its far more ambitious objective on this project was to acquire 6,000 km of 2D seismic data and attempt to process all of it through a complete, highly sophisticated processing flow including dip move out (DMO) and final migration (Fig. 2).

ONBOARD OPTIONS

Once ARCO's processing specialists began to investigate onboard processing systems, they found there were few viable options. In fact, several geophysical contractors stated that processing a survey this large through the full sequence ARCO proposed would be impossible with existing technology.

ARCO considered using its own Alliant 800 computer, which was used on certain land-based acquisition projects. It was already installed in a portable shelter that could easily be moved onto a ship. But its processing capabilities were too limited, and it had nowhere near the requisite capacity for the projected volume of data.

The second option was to install interactive processing workstations on the ship. The main question: Was it possible to obtain high enough quality and throughput from systems compact enough to fit onboard?

The concern was that this approach might require a prohibitive number of workstations or operating personnel,

ARCO evaluated an interactive system that seemed promising. Although it had an attractive user interface, it became apparent that even with multiple workstations the system would not have sufficient throughput to keep pace with acquisition.

Several months of processing onshore would still have been necessary before the interpreters could obtain the complete survey. This was simply not the solution ARCO was looking for.

The survey was scheduled to begin before the end of April, and it was already February. Time was running short. That's when ARCO first approached Landmark's office in Jakarta. Landmark proposed using a single interactive seismic processing workstation of its own, with a second system for 100% redundancy in case of hardware failure.

Only one workstation was considered necessary because the system uses multiple processing accelerators called "supercards," which exponentially increase throughput. In addition, its highly sophisticated algorithms were specially developed to optimize the accelerator hardware. As a result, the system provides both high quality and high throughput on one machine.

ARCO had Landmark benchmark the system by running a sample data set through the anticipated processing flow. Testing not only produced excellent processed sections but also indicated the system could handle the full volume of data to be acquired.

Two Landmark geophysicists-one expatriate and one Indonesian national-would man the system on each 12 hr shift for the duration of the survey; ARCO geophysicists would establish parameters and monitor progress.

ACQUISITION AND PROCESSING

On Apr. 19, Landmark installed two Sun 670 workstations on an acquisition vessel in Singapore. The workstations, disk towers, tape drives, and plotter were shock mounted and configured to prevent unnecessary vibration. No hardware failures occurred on the project, apart from some plotter problems.

During the 30 hr steam from Singapore to the Ardjuna basin, just off Jakarta, an ARCO geophysicist did the initial testing. Processing parameters were selected in real time, tested quickly, and modified to produce the best images.

The ship proceeded to the first priority area, where processing and acquisition parameters were fine-tuned before continuing. At this point, a software limitation turned up.

Because of limited time, benchmarks prior to the survey had focused on processing sample data through the complete sequence, not on the size of data sets. Many of the lines acquired in this survey exceeded 50 km in length, and at first the operators had to segment long lines to process them efficiently.

Landmark's development center in North America worked on the problem until modifications were made to handle much larger data sets. Updates to the software were sent out to the ship by satellite, and the problem was quickly resolved.

Early results indicated that the basin had somewhat deeper structures than ARCO's explorationists realized. As a result, two changes were made to the original plan.

First, the vessel's seismic source array was altered to obtain deeper penetration, recording 240 channels of data to 6 sec at a 2 ms sample rate. Second, interactive parameter testing indicated the need to do more complex DMO corrections than expected, to properly image the basin.

The decision was made to compute full 60 offset DMO instead of summing traces and doing 15 or 30 offset DMO. This decision effectively eliminated the redundancy of the second workstation, because full DMO processing is so computer-intensive.

While trace summing would have cut computer time for this step by 50-75%, it also would have compromised data quality. In addition, interactive velocity analyses were done every 4 km before and every 1 km after DMO, which increased the amount of human interaction required during processing as well.

Because of the initial software changes and the more complex processing flow, processing lagged behind acquisition (Fig. 3). To compensate, system throughput was beefed up by doubling the number of supercards in each workstation-providing, in effect, the computing power of a mainframe. After that, it was possible to run 14 or 15 jobs simultaneously, using both machines at full capacity.

Another factor affected the final amount of data processed aboard the ship. ARCO had expected acquisition to take at least 2 full months, lasting until the end of June, Due, however, to calm seas, the survey was completed almost 3 weeks ahead of schedule.

Acquisition rates averaged 142 km/day and exceeded 200 km/day on occasion. Because of this, the team managed to process only two thirds of the total 6,000 km before the ship docked.

However, processing rates-after software modifications and installation of additional accelerators - rose to an average of 140 km/day. It is clear, therefore, that had acquisition taken the full 2 months projected, all of the data would have been processed before the survey was completed.

In any case, ARCO's interpreters obtained 4,000 km of fully processed data that they might not have seen for another 8 months if processing had been done following acquisition.

CONCURRENT INTERPRETATION

About every 10 days, during personnel rotation, tapes of processed data were sent to Jakarta and loaded on interactive seismic workstations. Interpretation began during acquisition.

As a result, ARCO achieved one of the most important benefits of onboard seismic processing. The interpreters recognized that the programmed position of a certain line, based on the old data, was inadequate to properly image the subsurface on either side of an important fault. It was too close to the fault.

So they laid out two new lines on a basemap and sent it out to the ship. Those additional lines were acquired, processed onboard, and returned to Jakarta, where they too were interpreted before the survey was over.

Without onboard processing capabilities, it would have taken a year or more to accomplish the same thing--to process and interpret the original data, discover the need for additional lines, commission another vessel, shoot the required data, process it as well, and finally incorporate it into the initial interpretation. As it turned out, ARCO's interpreters obtained the seismic data they needed on the first pass, saving both time and the cost of a second survey.

After the survey was completed June 10, the workstations were installed in a dedicated processing center at ARCO's headquarters in Jakarta. By July, interpretation of data processed onboard at 4 ms had already favorably influenced a number of exploration well locations.

Nevertheless, the entire survey is currently being reprocessed at 2 ms to image its higher frequency content. Information from this large regional survey will be used for years.

The next interpretation phase will utilize the higher frequency data to better target shallow gas plays throughout the basin.

IMPLICATIONS

Because of this pilot onboard processing project, ARCO has significantly accelerated its natural gas development program in the Java Sea, helping to meet some of Indonesia's most urgent energy needs.

In addition, ARCO International is considering the use of full seismic processing-not just field acquisition quality control-on many other acquisition projects.

This was not the first time the company attempted onboard seismic processing. As early as 1982, a serious effort had been mounted but fell short of its goal due to technological limitations. Until now, there has not been a compact system with sufficient throughput and high enough quality to fully process the data onboard.

Many geophysical contractors currently developing interactive processing workstations assume that only front end parameter selection should be easy and graphical. But they continue to rely on mainframes and batch-oriented software to do the bulk of the processing.

Not only does this approach limit interaction between processing and acquisition, but it makes it virtually impossible for seismic interpreters to see the data in time to change the survey, if necessary.

Onboard processing has tremendous potential. But it is not a panacea. In regions where turnaround time is not as critical as it is in Indonesia, onboard processing may not be as urgent.

Also, in almost any other environment, processors work on multiple projects at the same time. Onboard seismic processing requires at least one highly qualified processing specialist to dedicate full time to a single project.

To avoid committing one of its own geophysicists for the duration of the survey, ARCO employed Landmark consultants to work on the boat.

Besides processing marine 2D surveys, ARCO is looking into other applications of onboard processing, including onshore 3D surveys and determining source configurations in very complex offshore areas.

The typical procedure for the latter is to have a boat shoot one seismic line several times with different seismic source arrangements. Then the data are taken to a processing center and the resulting lines compared to determine the best configuration.

An acquisition vessel then must be sent out again to shoot the survey, again followed by processing. An onboard system would eliminate the need for two trips to the field and speed up the exploration cycle.

CONCLUSION

Our experience in the Java Sea has shown that, in exploration programs with similar constraints, the advantages of onboard processing far outweigh any imaginable drawbacks:

Acquisition-to-drilling time can be drastically cut.
Costs associated with onboard processing are lower than comparable post-survey processing onshore.
Processing flows can be altered, if necessary, to accommodate new perspectives gained in the field.
"Concurrent" acquisition, processing, and interpretation enable explorationists to modify the survey design before the seismic vessel returns to shore, eliminating the need to reshoot inadequate surveys later.

Full onboard seismic processing has arrived. It may be in its infancy, but it has already proven its value in the field. Oil companies that take this first solid step toward more collaborative seismic acquisition, processing, and interpretation may find themselves in a better position as international competition grows more intense.