The Oil Drum | GAO Peak Oil Report Thread

117 comments on GAO Peak Oil Report Thread

Comments can no longer be added to this story.

| Show without comments | PDF version

Sam Foucher on March 29, 2007 - 3:01pm

From the chart, we can derive a probability density function:

Below is the cumulative density function (cdf):

There is 5% of chance (F05) that PO data < 2007, 50% of probability that PO data < 2034 (F50) and 95% of probability that PO date < 2114 (F95).

Comments can no longer be added to this story.

Halfin on March 29, 2007 - 3:37pm

Nice work! I'll repeat the reference to the survey paper on the advantages of combining forecasts:

http://faculty.fuqua.duke.edu/~clemen/bio/Published%20Papers/13.Combinin...

Abstract: Considerable literature has accumulated over the years regarding the combination of forecasts. The primary conclusion of this line of research is that forecast accuracy can be substantially improved through the combination of multiple individual forecasts. Furthermore, simple combination methods often work reasonably well relative to more complex combinations.

To me, this clean, simple averaging approach is far more likely to be a good guide to the truth than Stuart's efforts to read the tea leaves of every jog and squiggle in the Saudi production function.

Comments can no longer be added to this story.

Sam Foucher on March 30, 2007 - 2:19pm

Thanks for the article, it's a good reference.

Note that the maximum likelihood derived from the pdf above is between 2012 and 2017.

Comments can no longer be added to this story.

SamuM on March 29, 2007 - 4:10pm

Right tool, afaik, but GIGO.

Look at the data going in.

If they had included all the CERA/IHS/USGS rehash studies, they could have pushed the probability of <2034 to near zero.

What we need to feed is:

- independent (i.e. done without other studies data)
- different methodology (e.g. bottom up, HL, economic, loglets, whatnot)
- credible (i.e. now back of the envelope calculations)

...studies. Also, for balance measure, we would probably need also to include some earlier estimates that have been proven wrong (too early), because it is likely that some of the other late peak estimates are wrong too (assuming somewhat normal distribution, even if tailed).

Then we MIGHT get some useful data. Perhaps. Agree?

Comments can no longer be added to this story.

Asebius on March 29, 2007 - 4:35pm

Sounds like a plan. I'd love to see that done.

Comments can no longer be added to this story.

doctorbob on March 29, 2007 - 5:58pm

IMO, exactly right SamuM. A mean calculated from results produced with such disparate methodologies, has little value. Many of the studies (e.g. CERA) use data or assumptions that are demonstrably incorrect, or use methodologies that are irrelevant. Such results should be excluded from any mean that is calculated. Even then, a mean is of limited significance as most Contributors who have done detailed analyses and leading commentators (e.g. westexas) seem to be homing in on a narrow window of "nowish" to 2010.

The most interesting thing about this, is that it raises awareness of this probability amongst the MSM. It will be interesting to see how this develops and its effect on oil prices (now towards $66), esp. combined with Iran's antics with the captured sailors. It's almost as if they are trying to stir up antagonism and strengthen excuses for an attack later this year.

Comments can no longer be added to this story.

Halfin on March 30, 2007 - 2:08pm

I don't necessarily agree that you want independent data in this kind of averaging. If many experts agree with a particular line of analysis and write papers supporting each other, then it makes sense to take the fact that that is the widespread consensus into consideration. Anyone who fails to recognize that near-term Peak Oil is very much a minority view is deluding themselves. The averaging process should reflect that reality.

Comments can no longer be added to this story.

noutram on March 29, 2007 - 7:30pm

Always with the sexy graphs... :o)

Comments can no longer be added to this story.

cgron on March 29, 2007 - 7:37pm

Is it possible to redraw the pdf/cdf charts in 3D with depth representing which underlying data set was used (USGS, etc)?

In other words the analysis is too flat, equally weighting all the studies.

Comments can no longer be added to this story.

Engineer on March 30, 2007 - 7:44am

Great way to look at the data! I question your methodology a bit though. Shouldn't the forecasts spread over mutiple years only contribute a fraction to each year in the PDF? The way you plotted it, the total area in the plot represented by the studies with large ranges is disproportional to the area representing rest of the studies. Doesn't this overweight the studies with large ranges?

Comments can no longer be added to this story.

Engineer on March 30, 2007 - 8:25am

Note if you do the PDF/CDF the way that I suggested F05 is 2005, F50 is 2022 and F95 is 2053. I eyeballed the data off the chart so the numbers are probably +/- a year or so.

Can someone point me to a tutorial for posting charts?

Thanks.

Comments can no longer be added to this story.

Calorie on March 30, 2007 - 9:22am

You must first find a place to host your image (Google "Image Hosting" to find a number of free sites), then use the "img" tag. Here is the tag Khebab used for his first graph:

Note that if you are using Mozilla Firefox, you can select a section of rendered text and graphics and use the "View Selection Source" command to see the underlying HTML. You can use this to figure out how to do something someone else has done. You can do something similar in Microsoft Internet Explorer.

I don't think you can host an image at the Oil Drum unless you have special privileges (see "Personnel" block on the right-hand sidebar).

Comments can no longer be added to this story.

Sam Foucher on March 30, 2007 - 11:17am

hmm... good question. What would be the weight then?

Each study has the same weight for each year (i.e. 1/18). A study with a large time span (e.g. 2005-2140) would be basically non informative because it would then affect each year the same way (i.e. it would simply raise the minimum probability floor). Note that I have not included the studies that are just saying "After 2030", If I include them I get F50 around 2050.

IMO, studies that are including shale oil should have lower weight for instance because there is a low probability that this source will affect the peak position.

Comments can no longer be added to this story.