<DrumBeat: April 17, 2007 The Oil Drum DrumBeat: April 18, 2007>

101 comments on The Shock Model (Part II)

Comments can no longer be added to this story.

| Show without comments | PDF version

garyp on April 17, 2007 - 11:49am Permalink | Subthread | Comments top

The danger of 'reserve growth' is the nice mathematical form of it. Fig 5 seems to show an nice asymtotic 0.01 factor on any field, of any age, continuing forever.

However I'd suggest its right to question that.

Certainly over the years we have got better at identifying new additional reserves in existing fields. At the same time we have got better at ecking out the last drop from each field. The question is, once we have given a reasonable time for some certainty to develop in the size and shape of a field, is it right to still expect further growth over time? A field discovered in 1995 will be measured, probed and described within an inch of its life. Would we be right in assuming that it will continue to grow 1% per year, every year, even in 2030?

If you were to tailor a growth factor that tended to zero after 10 years for newer discoveries, what would that do to the predictions?

Comments can no longer be added to this story.
Sam Foucher on April 17, 2007 - 12:14pm Permalink | Subthread | Parent | Comments top

I agree. There are a lot of statistical wizardry and assumptions behind the estimation of the cumulative growth factors. Note that the last estimation by Verma (Modified Arrington) is less optimistic than the original Attanasi & Root used by the USGS. Strong apparent growth observed in North America is probably due to the way initial reserves (1P) are booked. Robelius, in his PhD thesis, notes that because initial reserve estimates are now much more precises, reserve growth is likely to decrease considerably in the next few years.

Re: If you were to tailor a growth factor that tended to zero after 10 years for newer discoveries, what would that do to the predictions?

it can be simulated by changing t_max from 95 years to 10 years in the code.

Comments can no longer be added to this story.
WebHubbleTelescope on April 17, 2007 - 10:51pm Permalink | Subthread | Parent | Comments top

Attanasi & Root applied a statistical technique to censored data without applying a correct survivability analysis to the complete data set:
http://mobjectivist.blogspot.com/2007/01/censored-data.html

This tends to overestimate reserve growth with assumptions on the improperly back-extrapolated data.

I have previously engaged in a heated debate over this issue at peakoil.com within the last year. This is basic "cooking the books" at the most egregious level. Bo-oh-boy, do the petroleum engineers get upset when you bring this subject up.

Comments can no longer be added to this story.
Engineer on April 18, 2007 - 12:10am Permalink | Subthread | Parent | Comments top

Alright, thats about the fifth time I've read this remark in various places and it exceeds my statistics background. Doh!

Any references to a good book or web site to brush up on censored data and survivability analysis? I prefer books ...

Thanks ...

Comments can no longer be added to this story.
WebHubbleTelescope on April 18, 2007 - 2:15am Permalink | Subthread | Parent | Comments top

I don't have a good ref handy, but you can get a start by applying a bit of common sense. Censored data is most commonly found in medical studies of lifespan, for example, how long people will on average live who take up tobacco smoking. The censoring comes about when we have limited time periods for the data collected; when this data is used beyond the censored region you generally run into problems. The tobacco studies would go on for 10 years or so, and the statisticians would collect enough data to try to project death rates in the future. The results become problematic when the extrapolator assumes a linear projection whereas in reality the actual data could vary non-linearly. You really can't go too far outside the censoring window, without having some pretty good assumptions to back the projections up.

This Wiki entry talks about discarding censored data properly in the context of a Kaplan-meier estimator for survivability analysis:
http://en.wikipedia.org/wiki/Kaplan-Meier_estimator

Censoring in reserve growth projections come about when a narrow window starting in the 1970's along with ongoing production data from wells discovered pre-1910 was used by Attanasi & Root to extrapolate reserve growth at the 80+ year mark. Even though they had no supporting data from reserve growth from 1910 to 1970, the spurts of reserve growth in the old wells during the 1970's allowed them to make cumulative reserve growths that appeared to keep tracking upward. But in reality those missing (i.e. censored) years could have shown any arbitrary amounts of growth -- including zero.

The only real way to deal with censored data is to get data from different periods of time, say from different experiments, and try to piece those together with any ingenious techniques you have at your disposal. Only then can you get a true empirical estimate of the growth. The problem with oil is we only have that window starting in the 1970's

From A&R, x-axis is the censored window, curves show raw reserve growth rates of USA oil regions of various vintage

Global warming researchers know how to piece together censored data. They can take recent CO2 samples, mix them with old ice core samples, and put together a fairly complete view of how CO2 varied over history. Oil people apparently don't have this kind of sophistication.

Comments can no longer be added to this story.
cynus on April 17, 2007 - 12:35pm Permalink | Subthread | Parent | Comments top

Looking at figure 7, it appears as if reserve growth has been proportionaly much smaller since 1980 than before.

Comments can no longer be added to this story.
WebHubbleTelescope on April 18, 2007 - 12:17am Permalink | Subthread | Parent | Comments top

That is a subtle effect as it takes a while to get the reserve growth to accumulate with time. Therefore the amount since 1980 looks smaller -- but it is also a lot less just because reserve growth amount is also proportional to the amount discovered and the amount discovered since 1980 has decreased substantially.

The big issue is how much reserve growth actually occurs. And we need to do a careful censored data analysis on the numbers, not the mumbo-jumbo that Attanasi & Root did for USA reserve growth estimates.

The censoring comes about since the sliding scale for extrapolation only goes back to 1970 or so (i.e. the censor date). Therefore, when you look at reserve growth additions for 1945, we have no idea what the reserve growth additions between 1945 and 1970 actually were. For all we know, these additions had been accumulating for quite a while before they got reported. This is the basic flaw with incorrectly using censored data.

Comments can no longer be added to this story.

101 comments on The Shock Model (Part II)

Comments can no longer be added to this story.

| Show without comments | PDF version