| View previous topic :: View next topic |
| Author |
Message |
pilferage
Intermediate Crude
Joined: Nov 21, 2004
Posts: 579
Location: ~170ft/lbs@0rpm (on my bike)
|
 Posted: Mon Dec 13, 2004 4:43 pm Post subject: EIA's Analysis vs. ASPO's Analysis |
  |
|
Just looking at the EIA's graph of production (pg 5)
http://tinyurl.com/4jb2q
the structure of their graph seems very atypical. Like we'll manage to somehow get most of the oil out and experience a rapid decline after. This hasn't been seen in the past
http://en.wikipedia.org/wiki/Image:Hubbert_world_2004.png
It's almost like they're implying we can somehow get a pre/post-peak ratio of ~2-3/1 as opposed to the historically based ~1/1 ratio of the logistics curve. What kind of iinfrastructure changes are they suggesting might occur? If we were to implement water/steam/co2 injection in every field in production right now and every field we bring into production later, would that induce this kind of curve?
Or am I correct in assuming we'll have a ratio of ~1>/~1< resulting from the current use of these methods since historically we haven't used them...
Here's ASPO's analysis...
http://www.peakoil.net/Newsletter/NL48/newsletter48.pdf
which seems spot on if you compare it to what's happened historically.
Last edited by pilferage on Sun Dec 19, 2004 4:46 pm; edited 1 time in total |
|
| Back to top |
|
 |
nero
Light Sweet Crude
Joined: May 22, 2004
Posts: 1422
Location: Ottawa, Ontario
|
| Posted: Mon Dec 13, 2004 5:19 pm Post subject: |
|
|
| Quote: | | Also, I'm lmfao @ the notion of the mean between the 95% and 5% chance being the expected value. You'd figure that if the usgs says there's a 95% chance there is 'x' amount of oil, they'd use that figure. But to include a mean of two seperate analysis, take their numeric mean of their two outputs (not even a weighed mean!), and state that it's the expected value is just asking for it! |
I don't believe you've got the right idea there. As I understand it the USGS performed a MonteCarlo Simulation to estimate the potential remaining reserves available. What this means is they made estimates of the various factors involved including an estimate of the precision of their estimate. They therefore instead of having a single estimate for each variable had a probability distribution for each variable. The then ran thousands of simulations in which the computer randomly created a set of values for all the variables (taking into account the probability distribution of each variable) and calculated the resulting potential reserves left. Each iteration would produce a different result and from the resulting data set a probability distribution was estimated. 95% of the iterations produced an estimate of total potential reserves available below the 95% number. 5% of the iterations produced an estimate below the 5% number.
In my opinion for this application a MonteCarlo simulation is of very little value. Why didn't they just add up the probability distributions for the different basins? My personal bet is because a MonteCarlo simulation sounds much more impressive. |
|
| Back to top |
|
|
pilferage
Intermediate Crude
Joined: Nov 21, 2004
Posts: 579
Location: ~170ft/lbs@0rpm (on my bike)
|
| Posted: Mon Dec 13, 2004 5:35 pm Post subject: |
|
|
Ah I see, so the mean value was done by the usgs using the same methods? I assumed from the graph notation that the eia had literally summed and divided by two... 
As for the MonteCarlo Simulation, if I follow you correctly, it seems as if they were running the probability disributions of the various probability distributions they had assumed...
so (unless something was really f'ed up) they had to come up with the same values.
Very pointless if you ask me  |
|
| Back to top |
|
|
tdrive
Heavy Crude
Joined: Jul 11, 2004
Posts: 350
|
| Posted: Mon Dec 13, 2004 6:58 pm Post subject: |
|
|
| Quote: | | My personal bet is because a MonteCarlo simulation sounds much more impressive |
Well, since the probability distributions are not analytically describable,
you cannot get away without a Monte Carlo. The problem is, that they
give the average of .95 and .05 as the expected result, what a load
of crap. Same as saying, I have a 50% chance of filling my gas tank
AFTER I pay at the gas station. Those who came up with that deserve
to get shot.
Cheers, |
|
| Back to top |
|
|
pilferage
Intermediate Crude
Joined: Nov 21, 2004
Posts: 579
Location: ~170ft/lbs@0rpm (on my bike)
|
| Posted: Mon Dec 13, 2004 7:41 pm Post subject: |
|
|
| tdrive wrote: | Well, since the probability distributions are not analytically describable,
you cannot get away without a Monte Carlo. |
You should be able to, you'd just have to take whatever value was associated with the p95 of each variable in your model and run just that one simulation. By running the Montecarlo they can prolly get a better grasp of the big picture (since they're running all the possible models) and might be able to draw some conclusions about how variables effect each other. But either way it's the same thing... |
|
| Back to top |
|
|
nero
Light Sweet Crude
Joined: May 22, 2004
Posts: 1422
Location: Ottawa, Ontario
|
| Posted: Wed Dec 15, 2004 8:15 am Post subject: |
|
|
| Quote: | | Well, since the probability distributions are not analytically describable, you cannot get away without a Monte Carlo. |
I'm not sure what exactly you mean by this, but after reflection I agree that a MonteCarlo simulation is appropriate. You do however come back to the garbage in garbage out phenonema (as pointed out pretty graphically by Campbell). |
|
| Back to top |
|
|
|
News
Discussions
Resources
Members
Forum FAQ
Search
Usergroups
Profile
Log in to check your private messages
Log in
