Zen and the art of oil exploration
Copyright ã 1987, 1999, Kevin T. Kilty, All Rights Reserved
Over a period of ten years Dr. Menard promoted a theory of oil exploration, and oil reserve estimation that grew increasingly cartoonish, and irrelevant. In one of the last books he wrote (1), he slanders the oil exploration business. His purpose became eventually to show how piteously inept oil companies are at finding oil, and at the same time proposes how to find oil on the cheap. One finds oil most economically, he insists, not by doing any geology or geophysics, but by drilling on a regular grid or drilling randomly. In other words, doing any research or thinking only makes the job of finding oil more difficult and expensive, one has to resort to an irrational strategy.
I am familiar with exploratory drilling on a grid. Not exploring for oil, but for gold. In my experience drilling on grids never works well; so, Menard's method seems suspicious to me on its face. His argument for the failure of exploration methods is much worse. Through a series of papers during the 1970s and early 1980s, he argued that oil exploration was being done very badly by the oil companies (2,3). The more popular the venue of publication, the more crude his line of thinking, and more sarcastic his attack. I am paraphrasing it, here, even though it may seem that I am parodying it.
Menard assumes that the probability of finding an oil field by just poking a hole anywhere is f/A; where f equals the area of the oil field and A equals the total surface area to explore. The probability of not finding an oil field is (1-f/A). Drilling two exploratory wells reduces the probability to (1-f/A)2; drilling n wells reduces it to (1-f/A)n. Thus, 1-(1-f/A)n becomes the probability of finding an oil field after drilling n wells. Now, Menard reasons, the size of East Texas field is 567km2 and the area of all sedimentary basins equals 4.7x10 6km 2; so f/A equals 1.2x10-4. The oil industry had drilled 300,000 wells by the time it found the East Texas field in 1930. Menard uses this for n in his formula, and determines that not finding the East Texas field before drilling all these wells is stupendously small Ė 2.0x10-16, to be approximate. Savoring his triumph, Menard says "...the reason for this bad luck is not wholly understood, but it seems clear that, in organized exploration by Western civilization, doing as well as pure chance may be something of an achievement."
If Menard is trying to win consulting contracts, he is not going about it well. I canít imagine oil executives taking well to being labeled so stupid, nor are they likely to take his exploration advice seriously(4). Menard proposes a type of Zen exploration in which one patiently drills holes relying on intuition, or a regular grid (a random program), blocks out the discoveries, and cashes royalty checks. Has Menard invented, with extremely modest effort, an exploration strategy more effective than thinking, research, and the expenditure of billions -- no trillions -- of dollars? The oil industry, keep in mind, employs lots of extremely smart people -- many of whom aren't Western.
This is a sorry model, but at least it presents a good example of how a very smart person can propose, and for a decade promote, a terrible idea. Anyone with a passing acquaintance with oil or mineral exploration, or probability, or statistics can dismiss the whole thing with one of the following three facts.
It is too simplistic
As evidence that Menard's model is too simplistic, consider the following facts. What constitutes an exploratory well? This concept has changed continuously from 1857 to the present time. Does the comparison of a present day exploratory well with Drake's discovery at Titusville make any sense? Of course not. Early oil men had no concept of exploratory work other than poking holes. In other words, the earliest oil-men were following Menardís random method. Except that they did drill where there were obvious indications of oil, like near oil seeps.
Many exploratory wells fail for technical reasons other than missing the field. For example, the drillers can screw-up the well. This must have happened often in the early days.
Exploratory wells may pass through an economic deposit without recognizing it. In fact, a failure can become a success at some later time. Thus, in the real world the a posteriori probability of drilling success can change from 0 to 1. Try modeling that with 1-(1-f/A)n! Recently there is evidence that some fields refill during production from non-economic deposits in the surrounding earth. How in the world can one model that with a single parameter?
Menard calculates probability using the size of the East Texas field. But the oil business did not begin with the concept of a giant field. Probably many initial wells wasted resources too carefully drilling the nearby area of a small discovery well. At what point does one consider these exploratory rather than development? Menard's model can't tell us.
Finally, an oil field ain't discovered by just drilling in the correct place. Oil fields have a limited extent in depth, also. So a person has to drill deep enough to reach the reservoir. The gross improbability of not finding East Texas within the first 300,000 exploratory wells ought to count only those wells deep enough to actually have reached the East Texas reservoir.
The model is not consistent
While Menard claims that drilling randomly succeeds better than organized exploration, he in fact presumes that the random drilling should only occur in sedimentary basins. So immediately the approach seems not random at all because we have to go out and map the sedimentary basins. Yet his statistical model states only that each well we drill improves our chances of finding Spindletop. Unless we go out and map the likely basins, we have to consider the entire earth as the area (A) in the model. This seeming handicap turns out to be convenient, though, because we now can drill into granite, drill into the sides of a Hawaiian volcano, or drill 6,000 feet through Greenland ice 9,000 feet thick, with the same chances of success as drilling in some place like say, East Texas!
Menard prefers a strategy of drilling randomly until we discover a field, then blocking out the remaining field. But this isn't a random strategy. It requires the recognition that oil occurs in pools. After all, we could waste a lot of resources stepping out too far, missing the pool, and then Menard would book this drill hole to exploration rather than development. Blocking out the pools requires some idea of how small pools might be -- information provided by thinking about and researching oil reservoirs. In other words by doing exploration.
Finally, the model doesn't explain many anecdotal observations regarding exploratory effort and success. For example, there is a statistic that Canadians love to quote, that they spend more on geophysics, and have better drilling success than Americans. A recent issue of Leading Edge describes a small oil field in the Sacramento Basin targeted with all sorts of geophysics (6). The first well, and eight of the next nine were successful. I consulted for a large gold mining company at one time that decided geophysics was a waste of time, and intended to drill on a grid to block out an ore body in Arizona. This company was big enough to spend a fortune and they did before they resorted to geophysics to help site boreholes more effectively with the tiny budget they had left.
Even Menard's own calculation shows suspicious behavior. A comparison of the actual discovery record against 10 Monte Carlo simulations of random exploration shows the actual record ranks among the top couple of simulations. Some simulations do a worse job by two orders of magnitude. In fact, considering the spread of simulated results, ten simulations hardly allow one to quantify the success of random drilling.
This model sets the wrong goal
While Menard figures that the tendency of industry to over explore certain popular targets explains their failings; he ought keep in mind that industry obeys economic forces beyond simply finding oil (6). An exploration strategy depends on other infrastructure. A purely random program that sends equipment skuttling across all the earthís lonely places is enormously expensive. What counts in oil exploration is not finding a giant field, but finding an economically feasible well. One has to pay for the exploration somehow. In this regard the actual history of oil exploration is far more pertinent than the anecdotal bad luck of not finding East Texas within 300,000 wells drilled. Actual exploration managed to build an enormous industry partially by plowing some resources into an effort guided by thinking. Random searches for resources often break the outfit that engages in them.
1/ H. W. Menard. 1986. Islands. Scientific American Books.
2/ H. W. Menard and George Sharman. 1975. Scientific uses of random drilling. Science 190, 337-343.
3/ H.W. Menard and George Sharman. 1981. Scientific American. 244.
4/ Menard cited many industry studies of calculating oil reserves and cites cooperation he received from the industry. It might be that by 1986 he and industry people were not getting along well.
5/ Leading Edge, January 1999.
6/ Menard actually refers to this in reference 2, but there is no way to include it in his model. Probably we should just add a single parameter to the initial assumption. The probably of success in a single instance of drilling is then, not f/A, but Sf/A, where S is a success coefficient that lies between 0 and 1. One could calibrate by using the first few decades of actual drilling experience, then analyze the remaining decades. This would show the oil companies to be spectacularly successful compared to random drilling.
7/ Menard and Sharman began this study as a means to calculate ultimate oil reserves. It would be interesting at this time to investigate how well their estimates have fared, and to see whether any alternative simple model would do as well.