The Oil Drum | Potential Impact of Cantarell's Decline on Mexico's Oil Production

78 comments on Potential Impact of Cantarell's Decline on Mexico's Oil Production

Comments can no longer be added to this story.

| Show without comments | PDF version

nero on July 14, 2006 - 12:33pm

The old adage that when you hold a hammer every problem looks like a nail seems appropriate here. The logistics curve is sometimes a useful forecasting tool but I think it should be used with discretion. Figure 5 demonstrates how inapropriate the logistic curve can be at times.

Comments can no longer be added to this story.

Halfin on July 14, 2006 - 12:58pm

Absolutely! The top half of figure 5 is the most unconvincing linear fit I've ever seen anyone dare publish. You've got dots making a flattish trend and then you arbitrarily take the last two and draw a line through them. You call that a linear fit? Come on! This part of the analysis is totally off base. You could plausibly draw a million other lines on that graph and get different results.

Comments can no longer be added to this story.

Sam Foucher on July 14, 2006 - 1:14pm

Look, everything is highly speculative here! Cantarell production has just peaked in 2003 and we have only two years of decline. However, engineers on the ground are saying that it will be a sharp terminal decline so no, you cannot draw whatever lines you want based on this information. Note also that the production levels predicted by the logistic curve on Fig 5. are the same as those given by PEMEX (Fig. 8) and Wood Mackenzie (Fig. 9): around 1.2 mbpd in 2010.

Comments can no longer be added to this story.

JCK on July 14, 2006 - 2:03pm

Look, everything is highly speculative here! Cantarell production has just peaked in 2003 and we have only two years of decline.

But the argument is that Hubbert linearization, in isolation (i.e., no USGS studies, no CERA data) allows to predict the peak in advance, right?

Two years ago would an HL fit have told us that Cantrell was at its peak? From the data shown, I would guess no.

Comments can no longer be added to this story.

Sam Foucher on July 14, 2006 - 2:21pm

The HL is not used here to predict the peak production (i.e. when we gonna peak and at what rate?), Cantarell has already officially peaked. The HL is used here to estimate the URR and the logistic decline rate (K) (i.e. how fast we gonna lose Cantarell's production). The resulting logistic curve is simply modeling Cantarell's decline post-2005.

Comments can no longer be added to this story.

JCK on July 14, 2006 - 2:42pm

I understand this. I am questioning whether the HL technique has any predictive value at all. If it didn't predict Cantrell's peak, why would the technique suddenly become predictive immediately following the peak?

While I like the elegance and simplicity of the method, the shown fits do not convince me of its utility.

Comments can no longer be added to this story.

Sam Foucher on July 14, 2006 - 2:55pm

why would the technique suddenly become predictive immediately following the peak?

I see two main reasons:

we know that a fast and terminal decline is expected (see the official chart on Fig. 8).
the predicted URR (17 Gb) is close to the official number (20.5 Gb)

I could have used something else that a logistic curve: a gaussian or a constant R/P ratio used by economist. Any monotonic curve that is delimiting an area close to 9 Gb will have done the job.

Comments can no longer be added to this story.

JCK on July 15, 2006 - 8:32am

Is there any reason to believe then that the HL method is better model than, for example, eyeballing the decline (based on the two post-peak points) and calculating the area under the graph to generate a URR figure?

Comments can no longer be added to this story.

Sam Foucher on July 15, 2006 - 9:15am

You can find some background on the logistic model applied to the modeling of ressource depletion in the following post:

Links to tutorial material on Hubbert Linearization

Comments can no longer be added to this story.

WebHubbleTelescope on July 15, 2006 - 11:35pm

It seems to work but no one knows why. I have a problem with understanding what the logistics K physically means. I think the real depletion rate for all Mexico is under 10% of available reserves using a more realistic rate model.

http://mobjectivist.blogspot.com/2006/03/mexico-oil-shock-model-part-2.html

Again, I don't think K makes any real physical sense.

Comments can no longer be added to this story.

westexas on July 14, 2006 - 1:36pm

"You could plausibly draw a million other lines on that graph and get different results."

This is true, so long as your "million other lines" fit within the constraints of an oil column of 825' thinning at the rate of about 300' per year.

Comments can no longer be added to this story.

westexas on July 14, 2006 - 2:00pm

In regard to the HL method, IMO the key point is that the HL plot of total Mexican oil production fits the same pattern that we have seen in other major oil producing regions--Texas; Lower 48; Total US; Russia; North Sea--and most recently, Saudi Arabia and the world.

The big fields tend to peak at the same time as overall regional production--Texas & the East Texas Field; Mexico & Cantarell; Saudi Arabia & Ghawar.

Based on Khebab's work, we published an article on Graphoilogy and the Energy Bulletin predicting that Saudi Arabia and the world were on the verge of the same kind of decline that Texas and the Lower 48 showed. Guess what? All of the available data to date show that our prediction was correct?

Instead of hypercritical discussions of how to linearlize the production in a field that is known to be declining, how about an acknowledgment of the success of the HL method, when applied to large producing regions?

Comments can no longer be added to this story.

westexas on July 14, 2006 - 2:02pm

Should be: "All of the available data to date show that our prediction was correct."

Comments can no longer be added to this story.

JCK on July 14, 2006 - 2:14pm

What happens with the Mexico data if we fit the HL data between 10 and 20 Gb cumulative production? How predictive was an HL fit from 1980-1990 of 1990-2006 production in Mexico?

Comments can no longer be added to this story.

kgmqt on July 14, 2006 - 3:04pm

From my understanding the HL method does not predict anything until peak is reached and steady state or decline has started - basically when only geology is the limiting factor in extraction rates. You can't use HL on a new developing field, or on a data that spans recovery methods (before and after N injection).

HL is used to predict URR after peak.

Comments can no longer be added to this story.

jkissing on July 14, 2006 - 3:37pm

in 1956 hubbert used it to predict that us lower 48 would peak in 1971, which is what happened. Since then aspo et al have tried to extend this success to the world, so far have repeatedly been too early.

Comments can no longer be added to this story.

Sam Foucher on July 14, 2006 - 4:06pm

the HL technique gives a credible production profiles for very few regions (Lower-48, Norway that I know of).

The following posts tried to assess the predictive power of the logistic curve fitting:

How Reliable is the Hubbert Linearization Method?
Bootstrapping Technique Applied to the Hubbert Linearization
How Reliable is the Hubbert Lin. Method? the world case

The HL fails when there is an abrupt change in the logistic growth (or decline rate) K which is the case for Cantarell, see this post from duncanK:

The Yibal Model - Does it Apply to Ghawar, Burgan, etc.?

Comments can no longer be added to this story.

westexas on July 15, 2006 - 8:32am

"Instead of hypercritical discussions of how to linearlize the production in a field that is known to be declining, how about an acknowledgment of the success of the HL method, when applied to large producing regions?"

FYI--Just to be clear, I was responding to Halfin.

Comments can no longer be added to this story.

GreyZone on July 14, 2006 - 3:16pm

/begin sarcasm alert
Yeah, Halfin, I guess the market would draw a rising vertical line through those two declining points and declare everything's ok, right?
/end sarcasm alert

Comments can no longer be added to this story.

ptone on July 14, 2006 - 1:54pm

I agree, "fitting" a line to two points is a bit of a joke. What an easy way to get and r of 1.00

Its not that the data don't deserve speculation and discussion, but using linear fit to come up with "hard" numbers, even if speculative, seems a bit disingenuous. Otherwise great post.

-Ptone

Comments can no longer be added to this story.

Sam Foucher on July 14, 2006 - 2:14pm

Ok, I surrender! The term "fitting" does not apply for two points, I should have called it "line drawing" or used three points instead :). Anyway, it would not have change the result significantly. Note also that the line gives an URR around 17 Gb consistent with the official estimate:

Cumulative: 11.6 Gb
Reserves: 8.9 Gb
URR= 20.5 Gb

src: PEMEX: Hydrocarbon reserves 2005 (Table A5, page 29)

Comments can no longer be added to this story.

78 comments on Potential Impact of Cantarell's Decline on Mexico's Oil Production

Comments can no longer be added to this story.

| Show without comments | PDF version

Search The Oil Drum with Google

Support The Oil Drum

Click the button to learn about supporting The Oil Drum with a secure donation.

Peak Oil Primers

Beware email scams!

Beware email scams claiming to be from this site. We do not have any job openings. If anyone contacts you about a job at The Oil Drum, do not reply to them, and definitely do not give them any personal information or send them money. Read more here.

“It is only through labor and painful effort, by grim energy and resolute courage, that we move on to better things.”

—Theodore Roosevelt