Wednesday, April 10, 2013

Seven Years More Data in the Hubbert Model


On Monday, I objected to some statements by Dennis Meadows that oil production had clearly peaked in the past and would now decline by half within twenty years.  He dismissed various possibilities for higher oil production from low grade sources as irrelevant.  To make my objections, I pointed out by looking at average growth rates, that the last five years data had delayed one's idea of peak oil (in the sense of average growth being zero) by around five years, and thus we can't have a very clear idea of when peak oil is exactly.

One objection I got in comments was that looking at average growth rates was a poor approach, and I should do something more sophisticated like look at the Hubbert model.  Fine; it happens that I did an extensive analysis of world production back at the Oil Drum in 2006.  Using the methodology of Hubbert Linearization, eg as exemplified by this picture:


I summarized the analysis at that time:
  • URR is 2250 ± 260gb
  • K is 4.93 ± 0.32%
  • the logistic peak is May 2007 ± 4.5 years
Here, URR is the total amount of oil that will ever be produced (in billions of barrels), K is a constant in the Hubbert equation that we needn't be concerned with here, and the peak date was in 2007, with a 95% confidence interval from the beginning of 2003 to the end of 2011.

So, I dragged out that old spreadsheet and updated the BP data from ending in 2004 to ending in 2011 - the last year they have published at present.  That gives us this picture:


The estimated URR is now at 2500gb, right at the upper end of the 95% confidence interval of my 2006 analysis.  So in seven years of additional data, during which time we used 210gb of oil, the estimate of total ultimate usage has increased by 250gb - more than we used in the meantime.  Again, this should give you serious pause in thinking that we know the peak with much precision.

The peak date is now in mid 2010 - ie it has moved later by about three years.  It is not out of the 95% confidence interval from my 2006 analysis, but I wouldn't now bet on it not moving out of it over the next seven years.  Instead, I'd be leaning on this caveat in my 2006 piece:
A last caveat. One of the major reasons for a linearization extrapolation to go wrong is that there's a big chunk of discovery that isn't even seriously started production yet. I do not think there are any such discoveries in the conventional oil world (the Caspian is quite small on the world scale, and I think deepwater is well under way). However, there are trillions of possible barrels of LQHCs (low quality hydrocarbons), such as tar sands, extra-heavy oils, coal-to-liquids, and then biofuels. That stuff can't be ramped in a hurry, but it will probably get ramped up eventually (depending on the climate wild card). The linearization is not taking account of those things.

26 comments:

Michael R said...

"One of the major reasons for a linearization extrapolation to go wrong is that...there are trillions of possible barrels of LQHCs (low quality hydrocarbons), such as tar sands, extra-heavy oils, coal-to-liquids, and then biofuels."

To be honest, I've never heard anyone talk about "peak low-quality hydrocarbons".

If that is the question, though, then yes, your argument is persuasive. We can't know when low-quality hydrocarbons will peak because we don't know yet how low the quality can go before it becomes not worth the trouble (economically, environmentally.

If, however, the question is whether or not the peak production of oil (stuff you can pour, that comes out of the ground) is shifting out unforeseeably, I don't think you've made that case yet.

Stuart Staniford said...

Michael R: Exactly - I wasn't trying to make that case. Instead I was pointing out that what Dennis Meadows is saying is not a reasonable extrapolation of the data we have up till now. He is making a statement about all oil, regardless of source. (And, really, that is kind of what society cares about).

Robert Wilson said...

I had a strange encounter with Dennis Meadows at an ASPO conference in Washington DC. I sat two feet from Dennis at a symposium that included Gail Tverberg and Nicole Foss. After the meeting I introduced myself to Dennis. He said "here it comes" turned and quickly walked away. In no way did I intend to criticize Dennis. I was an original fan of The Report to the Club of Rome and remain a fan. One of my all time favorite essays was written by the late Donella Meadows (on Thomas Jefferson). I have had brief but pleasant conversations with Gail and Nicole at ASPO meetings.

Luke Smith said...

Stuart said: So in seven years of additional data, during which time we used 210gb of oil, the estimate of total ultimate usage has increased by 250gb - more than we used in the meantime. Again, this should give you serious pause in thinking that we know the peak with much precision.

I read Hubbert's Peak to get an introduction to peak oil theory. I understand the concept but I wonder if there could be some problems with using the Hubbert curve. For one, I know indices, like stock market and price indices, will change based on the starting year of the index.

For example, just by updating your data, your peak date shifted from 2006 to 2010 after 5 years of new data. What if you started the Hubbert curve in 1970 or 1990, do you get different results?

Stuart Staniford said...

Luke: follow the link to my 2006 piece for much more on that kind of stability analysis.

Greg said...

On the one hand, it's interesting that URR has only moved out 250 gigabarrels. On the other hand, it shows the weakness of Hubbert linearization: it essentially assumes constant prices.

Jørgen Randers, one of Meadows's co-authors on The Limits to Growth, has more interesting things to say about the future in his book 2052: A Forecast for the Next Forty Years.

A 15 page PDF summary is available here.

Antonio Gallo said...

Simply, your previous peak oil prediction was wrong. In 2006 Jean Laherrère did a peak oil forecast. Conventional: about 2012, all liquids: about 2015. When someone ask him he replies: between 2010 and 2020

Stuart Staniford said...

Anthony: In the famous Campbell Laharrere piece in 1998:

http://dieoff.org/page140.htm

they forecast a peak for conventional and unconventional oil in the early 2000s. By now, it was going to be down about 10% from the peak and falling fast - indeed, as with Meadows now, they show it taking about 20 years to fall by half.

So I imagine Laharrere's evolution reflects a process of incorporating additional data (as it should).

yvesT said...

Stuart, I would have to dig deeper in the maths (and mines are a bit rusted, although could be revived I guess :) ) to make sure I don't say stupid things, but my undertanding is that :

- Deffeyes methodology allows to derive the URR from the production data (assuming a Hubbert curve)

- This is not the methodology used by Hubbert in 56, or by Laherrère now.

- That is, Hubbert first got an estimation of the URR from his knowledge of reserves, then on this URR he fitted a production curve.

- Laherrère is first using "creaming curves" using backdated discovery data (backdated standing by considering the real discovery dates, and not the date at which they are added to "reserves") to estimate the URR, and then fits the production curve on this URR.

In fact below an extract of one of Jean's emails regarding this :
"first the Hubbert (Deffeyes) oil production linearization is a poor way to estimate the oil ultimate and the best way is to use the creaming curve from cumulative backdated discoveries versus cumulative number of fields (or new field wildcats)
see my ASPO Brussels 2011 paper

-Laherrère J.H. 2011 «Backdating is the key » ASPO 9 Brussels 27 April http://www.aspo9.be/assets/ASPO9_Wed_27_April_Laherrere.pdf
Deffeyes had no access to backdated discovery data, only to production data
it is why he proposed this poor way to estimate oil ultimates.
Hubbert in his 1956 paper on US production forecast was not at all using the production so called Hubbert linearization, but the geological estimate being 150 Gb from his own and also 200 Gb being the highest estimate from a Delphi survey from the best world explorers carrried out by Wallace Pratt (one of the most famous explorers who stated that "oil is first found in a geologist's mind"

Unfortunately the aspo9 site seems to be down (not sure if temporary or not) for accessing the paper.

KLR said...

yvesT helpfully linked to Laherrère's latest update in the last post here at Early Warning. He sticks to an ultimate of 2200 Gb, notice.

A commentor at the piece here on Meadows wondered if tight oil were a phenomenon peculiar to the US. By now means; for one thing the formations in ND extend into Canada a bit; also there are huge resources of the same type spread around the world. Brian Wang wrote about the Paris Basin at his Next Big Future blog, for a start, and there are others, such as the Bazhenov in Western Siberia, which is 80 times the size of the Bakken. Full-on exploitation of these resources in the fashion of the US would completely upend estimates of peak oil based on conventional supply. Imagine Russia boosting production 20% in the span of 3 years.

Sam Charles Norton said...

How far is it possible/likely that production has been 'brought forward' through improved technology, ie a trade off between maintaining the plateau for longer at the expense of a steeper decline later? Conceptually that's a possible explanation.

Stuart Staniford said...

yvesT: Laherrere has used the linearization in the past, eg see http://www.mnforsustain.org/oil_forecasting_production_using_discovery_laherrere505.htm

Probably he liked it less over time - this is true for me too (I don't remember doing many linearization posts at all in the last four or five years). However, even at the start I only ever thought it was a very rough tool that had to be used in conjunction with other evidence.

Stuart Staniford said...

Sam:

I suspect it more has to do with the bringing on of more and more low grade sources (tight oil, tar sands, etc). There are of course enormous amounts of those sources, so it's not clear that the current rough plateau might not carry on for a long time. More on this in future posts I think.

yvesT said...

Stuart, went through Laherrère paper you linked, but I don't see where he is using linearization (determining URR from production history) in it, you mean for below ? :

"There is only one source of reserves data as WM does not report Romania fields. Oil and natural gas have a similar discovery pattern in oil equivalents. There is only one cycle and a second is unlikely. The ultimate seems to be around 8 Gb.

Figure 21: Romania creaming curve with U = 8 Gb for oil.



To check IHS values, the ultimate is guessed from production trend, which is only 6 Gb instead of 8 Gb "

Stuart Staniford said...

He shows linearizations for both Romania and Saudi Arabia (he calls them "P/CP" plots since I hadn't yet coined the "Hubbert LInearization" term when he wrote that paper :-). There may be others in there, but those two I remembered.

Greg T. Jeffers said...

Not that it is terribly applicable... but the equity (and commodity) markets do not seem to be terribly concerned with Peak Oil at the moment.

On the other hand, those markets do not seem to be terribly impressed with the prospects for Tight Oil/Shale Oil/Low Quality Hydro Carbon production.

Go figure.

buck smith said...

One thing most peak oil analysts and commenters miss about the shale oil boom in the US is how low costs are for oil companies in the US. We have terrific infrastructure and a very competitive market in oil field services. So the cost to drill a well is much lower than anywhere else in the world. That has a huge effect on the economics of shale oil production

Stuart Staniford said...

Buck - that's an interesting point. Is there any good data on that?

yvesT said...

Stuart, Ok, but he uses it more when nothing else is available, or as double checking (and for KSA takes URRs higher than the estimate from past production gives). In fact the first item in his conclusion :
"Reliable production forecasting needs reliable field discovery values."
Again I'm not a specialist in this, but intuitively relying only on production history greatly decreases the reliability of the projections, and this is especially true for countries where effective decisions on "not producing as much as possible" have been taken at some points.
Otherwise will try to check with Jean what was really included in his 98 "conventionnal and unconventionnal" projection (the red curve in the graph), as it doesn't seem exactly clear to me from the paper.

Antonio Gallo said...

But the point is: "Should The Last Few Years Have Updated Your Idea of Peak Oil?" or "Seven Years More Data in the Hubbert Model". Right? And the answer is: using Laherrere previous forecasts, no, the last few years haven't updated my idea of Peak Oil.

NGM said...

One shouldnt only look at one set of numbers...

The LQHCs provide much lower EROEI that traditional resources.

Thus you can have an increase in liquids production, but not necessarily have an increase in available energy for the rest of the system. At the same time the ecological factors are not counted in as expenses - even though they are expenses, but at a systemwide scale. Externalising costs leads to very misdirecting results.

Also We are having troubles with all sorts of ecological problems which will require we use larger amounts of energy to produce food or other necessary goods - since we cant "plunder and destroy" nature to the same degree and with same return quality as previously.

yvesT said...

A valid link for Laherrère paper above Laherrère J.H. 2011 «Backdating is the key » :
http://aspofrance.viabloga.com/files/JL_ASPO2011.pdf

Otherwise after an exchange with him, in the 98 paper, the red curve is in fact for "conventional", in the sense crude oil without extra heavy (Venezuela and canada tar sands).
So the "conventional and unconventional" text in the figure legend can be considered an error (or incoherency).

buck smith said...

Stuart,

My data is anecdotal based on working for Schlumberger as a field engineer in the 80s and work now as a contractor for an multinational services and equipment company. In general all multinationals service companies charge higher rates and have higher margins outside North America than in North America. There are several reasons for this but the big reasons are North American service market is very completive with many small players take business at lower rates. Also the infrastructure is really good- compare roads and pipelines in North America to Nigeria or Indonesia. This affects everyone’s costs with advantage to North America.
The way to get some data would be to look at day rates for rigs, prices for cement jobs, etc.

Danilo Bertocchi said...

Stuart
Jean Laherrère wrote as a conclusion on Feb. 4, 2013:
Conclusions
Production oil & gas data are unreliable and the UN should put as a priority to oblige every country to publish true and complete data on energy, in particular on fossils fuels which are the gift of humanity, because soon the world will suffer energy constraints.
The large increase of world population since 1850 is due to the progress in medicine and abundant cheap energy. We have entered the end of cheap energy and we need to save energy. But many do not want to change their way of life, count on a constant growth on energy, dreaming the renewable can replace fossil fuels.
It is a must to get the true data to show that fossil fuels will soon peak.
The present study is based on ultimate estimated mainly from extrapolation of discovery data coming from scout databases. Only three countries report reliable field data: UK, Norway and the Federal US. If most other countries follows their example, such study would be much more reliable. it is a shame that it is not understood in New York at the United Nations or in Brussels for the European Union.
With the poor present data it seems that world oil production will peak before 2020, Non-OPEC soon and OPEC around 2020.
OPEC oil export will cease before 2050.
Of course the present study is based on questionable assumptions and unreliable data, but anyone can look at the graphs and make his own interpretation or can challenge the data.
But I do not know any study showing as much as this paper.
I am sorry for the broken English and for being too long.
A shorter paper should be written after discussions within Clarmix or outside.
I will be glad to receive any comment and to discuss it. jean.laherrere@alsatis.net

Chris Reynolds said...

Stuart,

Hope you don't mind but I've linked to some of your graphics and posts in my latest mid monthly miscellanea post.
http://dosbat.blogspot.co.uk/2013/04/mid-april-miscellanea.html

Stuart Staniford said...

Chris - certainly not - thanks for the links!