ZONED AUTO-PICKING OF SEISMIC BOOSTS CAEX

Jay E. Valusek
Landmark Graphics Corp.
Houston

Mike Padgett, John Austin, R.L. Woodhams
Pennzoil Co.
Houston

Ian Page
Mobil Oil Canada Ltd.
Calgary

In less than 10 years, computer-aided exploration (CAEX) has revolutionized seismic exploration and field development. The increasing sophistication and ease-of-use of CAEX technology has yielded remarkable gains in productivity and quality.' Until recently, however, one critical aspect of seismic interpretation-"picking" subsurface horizons line-by-line-has remained essentially unchanged from paper to computer. Because of this, geoscientists often interpret only a fraction of their data.

With the advent of "zoned auto-picking" software in late 1989, interactive seismic interpretation has entered a new era: for the first time, oil companies that have invested huge sums of money in acquiring and processing three-dimensional (3D) surveys will be able routinely to incorporate a// of this data in their final interpretations. Two case studies, based on early use of zoned auto-picking, demonstrate the quantum leap in productivity this new technology offers. But first, a little background.

"PICKING" SEISMIC DATA

Manual picking. In the search for new reserves, oil companies have acquired increasing numbers of seismic surveys. Two-dimensional (2D) surveys may be small, with widely-spaced and randomly-oriented lines. 3D surveys tend to generate hundreds of seismic lines in closely-spaced rectangular grids. Thus, a 3D "volume" contains huge amounts of vital information. But all this data must be interpreted, all the horizons "picked" or "tracked," before maps and other images of the subsurface can be created. Traditionally, picking was done manually by drawing with colored pencils on paper, one seismic section at a time-an incredibly tedious process. Today, there are several other picking methods available as well (Fig. 1).

Auto-picking. In the early 1980s, interactive CAEX workstations gave seismic explorationists the ability to manage and analyze 3D data more effectively. While interpreting seismic lines was still accomplished by viewing and picking one line at a time, there were now two ways to do it: by drawing across the screen using a mouse and cursor, or by clicking the cursor at selected points on a reflector and letting the machine pick the rest. The latter, the first type of automated picking, represented an incremental increase in productivity over manual picking. Today, this is what most CAEX users mean by "auto-picking." Unfortunately, because of the number of lines in 3D surveys and the time it takes to look at each one, interpreters rarely pick every line. It could take months. Often, they must settle for every second or fifth line, leaving 50 to 80% of their 3D data uninterpreted. Obviously, something more is needed if companies are to take full advantage of the data they have acquired.

A more advanced auto-tracker, introduced in the mid-80s, attempted to fill this need. It allowed users to pick all of the 3D seismic data in a survey without having to look at every line-another increase in productivity. But the software required a specialist to run, and interpreters lacked real control of the picking. It tracked the whole survey in only one way, regardless of variations in geology and data quality. It didn't let users skirt poor data, fault zones, or salt domes. Nor could they focus on selected areas of interest. As a result, auto-picking generated numerous unreliable picks, and considerable time was spent editing or deleting horizons. This offset much of the initial gain in productivity. Interpreters began asking for greater flexibility, accuracy, and ease-of-use.

Zoned auto-picking. The result is "zoned auto-picking" (ZAP), which has already begun to have a significant impact on interactive seismic interpretation around the world. Interpreters now have a choice: they can track an entire survey, or subdivide it into zones of interest and track each zone in a different way. This flexibility allows explorationists to tailor the program to their needs, and makes ZAP useful even in areas with discontinuous reflectors.

There are two types of zones-inclusive (within which the program tracks a specified horizon) and exclusive (where no auto-picking occurs). The interpreter defines these simply by outlining polygons on a control map. To track a horizon, the user must input at least one "seed point," which then explodes in all four directions within the 3D data volume until it reaches the boundaries of the polygon-like tossing a stone into a still pond. But the number of control points or input lines is also up to the interpreter, and depends on the amount available or deemed necessary to reliably track each zone or fault block.

Interpreters have the option of tracking in a "turbo" or "iterative" mode. Turbo searches seismic traces adjacent to seed points for similar amplitudes, then races to the next trace without double-checking. This mode is appropriate for unbroken reflectors such as regional markers. For greater precision, the iterative mode cross-references every pick with the previous trace (or, adjacent traces if necessary) before making its final selection. Users also can specify (on a scale of 1-10) the degree of amplitude similarity they will accept. Therefore, the iterative mode works even where data quality is lower and geology more complex.

Unlike earlier methods of auto-picking, ZAP represents an exponential increase in productivity and accuracy. The following case studies from geographically and geologically distinct regions illustrate these benefits.

CASE STUDIES

Pennzoil Co., Gulf of Mexico. The giant Eugene Island Block 330 field, offshore Louisiana, is one of the largest oil-producing fields of the U.S. outer continental shelf, encompassing parts of seven blocks.' Block 330 alone has four platforms and nearly 80 wells. Since spring 1989, Pennzoil has been reevaluating a 3-D seismic survey of the block for detailed development mapping using CAEX technology. An aggressive drilling program has been outlined for 1990-91.

The survey consists of 175 lines by 410 traces of high quality 3-D data. Over a period of months, seven key producing horizons had been picked on approximately every fifth line, or 20% of the data. In early November 1989, Pennzoil obtained zoned auto-picking software and decided to input all of these lines to the program, hopefully to obtain the highest resolution maps ever done on Eugene Island Block 330.

Setting up the seven ZAP runs took only 30 minutes. Because 330 is split by a major NW-SE trending growth fault and relatively complex minor faulting, each horizon was subdivided into appropriate zones. Input lines as well as inclusive and exclusive polygons for the JD sand, an oil and gas producing reservoir, are shown in Fig. 2. All seven horizons were picked in the iterative mode, using correlation levels between 2 and 5, requiring the program to match traces with a high to moderate degree of similarity.

The program ran overnight and picked a total of nearly 130,000 traces. Each horizon took less than two hours to pick. Output horizons, such as the JD sand, produced detailed, nearly continuous seismic contour maps (Figure 3). Areas the program was unable to pick (blank spots) were minimal

As a result, a high quality interpretation of 80% of the 3D data required only thirteen hours, during time the workstation normally sat idle. Working full time, an interpreter probably could have picked the original 20% in a few weeks. Therefore, on this project, ZAP increased productivity several hundred to several thousand percent. Other key benefits included more accurate amplitude extractions and more cost-effective usage of computer time.

Mobil Oil Canada Ltd., Western Canada. The Sierra field in northeastern British Columbia has been producing gas since the late 1960s. It is a large bioherm consisting of a dolomite productive zone in excess of 270 ft in places, capped by a limestone over 850 ft thick, surrounded by shale.' On seismic data, sudden changes 'in dip over the reef make some horizons extremely difficult to pick.

In early 1989, Mobil Oil Canada acquired a 26 square mile 3D survey of the Sierra field with the goal of mapping the size, orientation, and structure of the reef for more efficient reservoir management.

Processing of 289 lines by 468 traces of excellent 3D data was completed in mid-October. Originally, six months had been budgeted for interpretation and mapping of ten horizons across the survey, using one of Mobil's geophysical workstations. Due, however, to internal decisions related to gas marketing, the deadline was unexpectedly moved forward to mid-December. In order to complete the project in two months, it was streamlined to three key horizons. Estimated time for the interpretation phase (picking every line) was 80 to 100 hours during the first 4 weeks.

Exploration Innovations (El) is a Calgary service company that performs most of Mobil's data loading. After learning they were also a beta test site for a new zoned auto-picker, Mobil decided to carry out the interpretation on one of El's machines. Interpreting three horizons on approximately every 20th line and a few crosslines (with a greater number over the reef itself) took part of 2 days.

These "seed lines" were input for ZAP to track in the iterative mode at a correlation level of 10, which requires only minimal similarity between adjacent traces. According to Russ Sagert, Exploration Innovations' system manager who ran the program overnight, it took approximately one hour and 45 minutes to auto-pick each horizon. At slightly over 135,000 traces/horizon, the software picked more than 77,000 traces/hr. The next day was spent reviewing picks and performing minor editing.

Total time spent on the interpretation phase-rather than 4 weeks-was only 14 hours. ZAP saved Mobil about 75 hours of work on this project alone.

The time saved was reinvested in fine-tuning the horizon interpretations and working on other ongoing projects. Mobil Oil Canada later acquired the zoned auto-picker and began using it on different types of production projects, with more complex geology, including a large 3D survey off the east coast. There, a single horizon can take several months to pick using the standard auto-picker, forcing interpreters to ignore horizons they would like to map.

With ZAP, preliminary results have been very encouraging. Besides use in production, this technology will likely also have an impact on time-sensitive exploration 3D plays in the near future.

CONCLUSION

In spite of its relative youth, zoned auto-picking has already set new standards in speed and user-friendliness' The advantages it offers oil companies are: more efficient use of experienced interpreters, interactive workstations, and 3D data. Within 6 weeks, for example, during nights and weekends, Pennzoil auto-picked enough traces of 3D seismic to circle the globe twice.

For some companies, ZAP may provide the impetus to adopt CAEX for the first time. By vastly improving productivity, ZAP also frees time for geoscientists to employ other interactive techniques to improve the quality of their interpretations and drilling recommendations. Look for this to become a standard part of the computer-aided explorationist's toolkit.

REFERENCES

Valusek, J.E., "Reports from the front lines: the global revolution in geophysical interpretation," Oil & Gas Journal, October 30, 1989, pp. 78-82. Valusek, J.E., "Computer-aided exploration: How well does it work?," Petroleum Management, March/April 1989, pp. 47-51. Valusek, J.E., "2-D seismic workstations help to site wells, evaluate leases," World Oil, September 1989, pp. 43-49.
Holland, D.S., Nunan, W.E., Lammlein, D.R., and Woodhams, R.L., "Eugene Island Block 330 Field, Offshore Louisiana," in: M.T. Halbouty, ed., Giant Oil and Gas Fields of the Decade 19681978, Amer. Assoc. Petrol. Geolo. Memoir 30, 1980, pp. 253-280. Holland, D.S., Lammlein, D.R., and Leedy, J.B., "Eugene Island Block 330 Field: An Update," presented at AAPG Annual Convention, Houston, Texas, March 2023, 1988. Holland, D.S., Leedy, J.B., Lammlein, D.R., "Eugene Island Block 330 Field, Offshore Louisiana, U.S.A.," in: E.A. Beaumont and N.H. Foster, eds., Treatise of Petroleum Geology, Atlas of Oil and Gas Fields, Amer. Assoc. Petrol. Geol., in press.
Cullins, J.F., Lake, J.H., "Sierra Reef Complex, Middle Devonian, Northeastern British Columbia," in: H.H.J. Geldsetzer, N.P. James, and G.E. Tebbutt, eds., Reef, Canada and Adjacent Areas, Can. Soc. Petrol. Geol. Memoir 13, 1989, pp. 414-421.