I was chatting to somebody at the Anaheim SPE ATCE who made a connection between risk management as applied to oil and gas exploration, decision support and the subprime fiasco. I pricked up my ears because I had just come out of a good talk* on the topic of how too much of our effort was spent on ‘risk’ evaluation and management and not enough on decision support.
The rapprochement between oil and gas risk analysis and subprime is both instructive and rather sensitive. For a while, risk management as applied to oil decision support has basked in the reflected glory of its use by financial institutions. After all, if the banks used it, it has to be valid. This is the old grass is greener argument. Since the French civil service is a big user of graphology, or because the US police once used a pendulum dowser, these techniques must be valid (not!).
One risk managers’ argument for subprime modeling failure makes reference to ‘outliers,’ unusual events that are ‘poorly represented’ by classical Monte Carlo modeling. Then ensues a sales pitch for a ‘new improved’ modeling technique, applicable both to financial services and oil and gas.
But wait a minute. The subprime crisis was not a modeling fault. The problem of subprime was understandable with no fancy models by any dummy who had access to the facts. The models weren’t thrown off by an ‘outlier.’ It was just that the Ponzi scheme bubble burst! A boom and bust phenomenon that has been around for centuries. In essence, it was the use of a steady state model that failed in an explosively dynamic situation. In other words, it was the wrong model.
But now that risk modeling in financial services has come unstuck, should we be looking at oil and gas risk analysis? Judging from the debate at the SPE and the ongoing discussions of all things risk and reserves related, it appears that not only we should but we do. One field where modeling is getting a big push at the moment is in reserves evaluation. A poster paper** argued that ‘simulator models constrained by a reasonable history match are likely to produce a reasonable recovery factor in a mature well or field.’ Here there is a push from industry that the SEC should admit more model-based estimates in financial reporting.
One reason that this seemingly innocuous notion is so hotly debated is that the history match itself is a little shaky. In some circumstances, like well understood, mature fields, it has obvious merit. Early on it is much harder to defend. The P10, P50, P90 categorizations lend spurious veracity to what are simple guesstimates. Late in a field’s life, when the field’s geometry, the rock’s wettability and a host of other parameters have been constrained by measurement, modeling can be used with more credibility.
So in general, oil fields and hence reserve numbers will fall somewhere along a spectrum of ‘unmodelable,’ experience-based guesswork, with a large degree of exposure to the sort of model ‘blow up’ that the subprimes experienced, to a hopefully more stable kind of modeling in the mature field where surprises are less likely. Quite how this spectrum of models can be rolled up into SEC reporting is beyond me.
Which brings me to my third modeling exercise of the month and an even bigger debate than subprimes—global warming. George Chilingar’s talk at the SPE was guaranteed to pull in the crowds. According to Chilingar, ‘Humans are not responsible for global warming,’ it’s all down to variations in the sun’s output (more on page 6). Chilingar’s thesis, backed by lots of back of the envelope calculations on ‘adiabatic this’ and ‘upper atmosphere that,’ boils down to a simple idea. That human energy output is and always will be a mere pinprick in comparison to the amount of energy that reaches us from the sun. So burning fossil fuels ‘makes no difference.’
This argument has been largely debunked***. Global warming is not about humans heating the earth up, but about greenhouse gases keeping the sun’s heat in. I find it strange that Chilingar, instead of taking the opportunity of debunking the debunkers and moving the debate along, just used the SPE to reach out to a new audience with his original spiel. But his argument was well received by the audience of petroleum engineers—especially so when a speaker from the US Environment Protection Agency offered strong backing to Chilingar’s position. This got me curious, was this official EPA position? Not exactly. Bottom line for the EPA is ‘Human activities are changing the composition of Earth’s atmosphere. Increasing levels of greenhouse gases [...]are well-documented and understood [and are] largely the result of human activities such as the burning of fossil fuels.’
At the risk of repeating myself I would like to point you back to my October 2003 editorial where, back from another ATCE , I located a great modeling quote from Hugh Gauch’s book—Scientific Method in Practice**** in that, ‘a model cannot be judged from its performance in predicting the data that were used to fit it in the first place’. What exactly does this mean for history matching, for oil and gas or climate change? I’m not sure I know, but perhaps we should ask ourselves if we really have the right model from time to time, rather than worrying too much about the minutiae of a particular probability distribution.
* SPE 109610, ‘Decision making in oil and gas.’ Eric Bickel (Texas A&M).
** SPE 110066 ‘Simulation in reserves—a guide for SEC/SOX compliant reporting.’ Dean Rietz,(Ryder Scott).
*** For debunking of Chilingar and other denialists see royalsoc.ac.uk/page.asp?id=6229.
**** ISBN 9780521017084.
© Oil IT Journal - all rights reserved.