[Graeme]

This was posted at TOD recently. I've reposted it here because I know that some of the mathematicians/engineers viewing this site will be interested. His conclusions are also of general interest.

Modeling Oil Production to Estimate URR - Saudi Arabia, Kuwait and World

$this->bbcode_second_pass_quote('', 'M')ore recently there has been much discussion regarding estimating a country’s URR using HL, Hubbert’s linearization, and especially the URR of the Kingdom of Saudi Arabia (KSA) and its potential oil production decline due to the depletion of the Ghawar oilfield (Ghawar reserves update and revisions by Euan Mearns and Depletion Levels in Ghawar by Stuart Staniford). To get a better understanding of the logistic function and its role in peak oil analysis, I decided to combine the concept of least squares with fitting the sum of multiple logistic functions to the oil production history of KSA as suggested by Khebab.

The least squares approach would provide a best fit to the country’s oil production and would address my curiosity to assess the quality of the results this method of analysis would produce. One of my objectives was to find another methodology that would complement HL and at the same time provide further insight into those situations that are difficult for HL.

[MD]

Once again a great modeling discussion at TOD and definitely worth your review time.

I remain in the "plateau now" camp, even if KSA URR turns out to be 260 Billion barrels simply because above ground factors in combination with the rapid depletion of sweet light will continue to constrain new production efforts across the board.

New production will never again surpass depletion. Systemic energy starvation is taking over.

[Qolio]

I'm sorry, but imho that whole analysis is nonsense. Like many people in TOD's discussion already pointed out, he's overfitting the data. You can fit any data if you have enough variables, but at the same time the predictive value goes down to zero. With this many variables the extrapolated curve is instable, and it's only pure luck that the URR estimates are consistent with some previous predictions. It would take only couple of small changes in last few years' data and the amount of reserves left would suddenly be practically infinite or near zero.

No good enough reason is given why the sum of several logistic functions can be used for fitting, and why the particular numbers of functions were used. It seems like the assumption is that there are multiple "subphenomena" (like individual oil fields or something), each of which individually follows the logistics curve. If this really is the case, then they should be isolated and analyzed separately.

[whereagles]

Unstable? Why? Care to plug some numbers into the equations and show that out explicitly?

[Qolio]

$this->bbcode_second_pass_quote('whereagles', 'U')nstable? Why? Care to plug some numbers into the equations and show that out explicitly?

[whereagles]

thx. I'll try and check your fits out this week-end

[whereagles]

Don't know if you're still there, but I finally got like 5 mins to look at the stuff

So basically what you're saying is: "I can make curves that fit the data beautifully, whose future behaviour ranges from doom to boom", right?