Resource Forecasting for the Geologically Challenged
Estimating the amount of crude oil resources seems to be a popular activity nowadays, but often the results of the various studies are not in agreement with each other.
Several independent researchers estimated the total planetary endowment, or “ultimate recoverable resource” (URR), as around 2 trillion barrels (ASPO 2006). Others, instead, propose larger numbers: a study by the United States Geological Survey (Wood and Long 2000) proposed an URR of 3 trillion barrels as the most probable value. A recent study from Cambridge Energy Research Associates (CERA 2006,1) proposes “4.82 trillion barrels” as the ultimate world endowment.
There exist even larger estimates, and Julian Simon (1981) made the curious claim that if you can’t measure exactly how much oil there is, you can’t deny that it is infinite (even more curious is that his claim had so much success). More recently, a vociferous crowd started claiming over the internet that crude oil is infinite because it is continuous re-created in the earth’s mantle by mysterious abiotic processes.
We may safely dismiss the claims of infinite oil as the result of a combination of wishful thinking and incompetence, but there remains the problem of the large discrepancy in studies that are claimed to be done using scientific procedures and to be based on reliable data. You would think that, as we keep accumulating more data on the world’s reserves, estimates should converge on a progressively better value. Instead estimates appear to be spreading more and more over widely different values.
According to ASPO, since we have extracted so far about one trillion barrels, we are left with, roughly, 1 trillion barrels. However, CERA says (2006, 2) that we are left with “3.74 trillion barrels.” Such a large difference, more than a factor of 3, has effects in how we picture our future. According to ASPO, the peak of oil production will take place within the first decade of the 21st century. The CERA estimate moves forward the oil production peak of at least two decades. Someone has to be badly wrong here; someday we’ll know which estimate was right. But, right now, how to decide?
Estimating geological resources is a job for the specialist. It is an arcane and difficult field for those of us who haven’t spent a sizable fraction of their life studying geology and not just geology, because the field also involves complex statistical treatments of the data. If you are not an expert, the task of reviewing in detail these studies is nearly hopeless.
Still, there are general considerations that can help even the geologically-challenged among us in understanding the reliability of the estimates. Some people have noted the suspicious tendency of CERA of giving too many decimals for highly uncertain quantities. That is already a hint that something is wrong, but this is hardly enough to dismiss the CERA study study entirely.
More effectively, you can ask yourself a simple question: how did previous resource estimates fare when compared with reality? Here, you can learn a lot from history and you don’t need a PhD in geology to judge by yourself.
One of the most studied historical cases of resource forecasting is that of the USA. In 1956, Marion King Hubbert estimated that the crude oil URR for the US lower 48 states would be between 150 and 200 billion barrels (Gb) and that the peak in production would occur within the interval 1966-1971 (Hubbert 1956).
Some years later, the USGS produced a number of studies which estimated an URR of around 600 Gb for the same region and concluded that there would be no limits of production for the foreseeable future (Zapp 1962, Sterman 1988). The peak for the US-48 did occur in 1970 as Hubbert had predicted and the USGS was wrong of a factor of three in their estimate. How could it happen that a government agency that is expected to be staffed with serious professionals could get it such an off the mark forecast? It is a story that has been analyzed in detail by Sterman et al. (1988).
The paper by Sterman et al. reports a complete story of the estimates for the US-48 region. As we see in the figure below, the estimates started low at a time when little was known about the geology of the continental United States. Then, as new data became available, the estimates increased, “peaking” a few years before the production peak.
Figure: Estimates of petroleum resources in the United States (Sterman, J. D., Richardson, G. P., & Davidsen, P. 1988)
Today, the cycle of oil extraction in the US 48 states is nearly completed, and we can reliably estimate the amount of oil that will be extracted as around 200 Gb (from Jean Laherrere, private communication).
Several of the estimates were way too high and Sterman attributes the problem in part to subjective bias; as we can read in the paper:
“Though ostensibly superior to the Hubbert method because it involves the use of disaggregate, primary geologic data, the analogy method (used by the USGS) actually involves a high degree of judgment, extrapolation of past trends, and educated guessing.
The simulation results suggest the substantial overshoot of the estimates was a consequence of systematic biases intrinsic to the method, biases that persist even when a high degree of perfect information is assumed. In a sense, the superior performance of Hubbert’s method can be attributed to its “mechanical” or formal character.
Precisely because it relies on only a small number of data inputs, and processes these inputs in a straightforward manner, it does not create the opportunity for subjective judgmental biases to creep into the forecast. “ (emphasis added)
These “subjective judgmental biases” are the likely result of the psychological pressure on the people performing the forecast that tends to make them choose optimistic data rather than the reverse. According to Sterman, badly overestimated forecasts are also likely to come when a producing region has been thoroughly studied and it is possible to perform a detailed “disaggregated” analysis.
CERA’s claim of operating using a “rigorously bottom-up procedure” makes their method very similar to the ones that the USGS used in the 1960s. For this reason, there is a good chance that their recent forecast for the world’s crude oil URR is overestimated. The CERA estimate may well turn out to be wrong by a factor of three, just as some of the USGS estimates of 1962 for the US-48 were. By analogy, we might even conclude that the fact that such a large estimate has been reported is in itself an indication that the global oil peak is close.
Resource estimates seem to be subject to something that reminds the uncertainty principle in quantum mechanics. When you have the perfect estimate, it is also perfectly useless. For instance, in 2005, the last French coal mine was closed. It was the end of a cycle of extraction that had lasted almost three centuries. Now, we know exactly how much extractable coal there is in France but it is a datum that is not very useful: that amount is zero. But when you are still extracting a resource and you need estimates, you must take into account that estimates are not only uncertain, but that their uncertainty does not simply go down as you gather more data. Estimates, actually, follow a cycle; they start too low and they get too high before settling to the right value. Much confusion could be avoided remembering that this tendency exists but, unfortunately, people tend to make the same mistakes over and over. Those who perform resource estimates without knowing the history of geology are condemned to repeat it.
Published online on www.aspoitalia.net on Nov 25th 2006
Association for the Study of Peak Oil and Gas and Department of Chemistry, University of Firenze, Italy.
ASPO 2006 (accessed) www.peakoil.net
CERA 2006 (2)
Hubbert, M. K. 1956, Nuclear Energy and the Fossil Fuels. Drilling and Production Practice. Washington: American Petroleum Institute, 7-25.
Simon, J. 1981, “The Ultimate Resource” Princeton University Press, Princeton.
Sterman, J. D., Richardson, G. P., & Davidsen, P. (1988). Modeling the estimation of petroleum resources in the United States. Technological Forecasting and Social Change, 33, 219-249, via MIT.
Wood J., Long G., 2000 Long Term World Oil Supply: A Resource Base/Production Path Analysis, via USEIA.
Zapp, 1962 A. D. Future Petroleum Producing Capacity of the United States. USGS Bulletin 1142-H.