Forecasting the Future

Conventional wisdom tells us forecasting the future is extremely difficult. Humanity must have millions of experts in the field. Some use simple methods (a crystal ball comes to mind), others use past history and extrapolate tendencies, and still others build very complex mathematical models, some of which require years of development and are run in giant super computers.

After reading another set of arguments about global warming, whether the world will hit 4 degrees C by 2100, sea level rise by 2 meters and such niceties I decided to take a simple look at USA oil production forecasts and see what I could learn about forecasting in general…

I started by downloading the USA´s oil production history. Let me clarify....In recent years we have seen the US “oil production” figures get mixed up with “condensate production”. You see, condensate is a liquid produced from gas reservoirs. Condensate is mostly light hydrocarbons which condense as the reservoir gas moves from the rocks to the surface production facilities. These molecules condense because the surface happens to be cooler than the reservoir (there are other subtle influences but I´ll skip the details).

USA condensate production has been surging because natural gas wells are being drilled to target shales containing natural gas with a high condensate ratio. This is a natural result of the lower natural gas prices which result from the excess supply over demand (an excess supply caused by the enthusiasm of companies going bonkers and drilling and fracking too many wells in recent years).

I´ve gone over the condensate versus oil issue because I see lots of production curves which mix products (they even go as far as adding biofuels to the total), and this gives the reader a warped idea of what´s going on. So I´m going to stick to show you what happened in the past, and what may happen in the future as far as OIL production is concerned:

The figure below is sort of complicated. It shows historical oil production, and some forecasts. I decided to show you what happens when one extrapolates tendencies using simple curve fits. 
My complicated graph

First I show you what happens if one were to take production until 2001, and crank up the equations…the extrapolated forecast shows USA production would have gone to hell very fast (that´s the dark dots falling like a rock). 

Then I show you the same technique applied to the full history through 2013…the extrapolated forecast goes bananas and says the USA will be a super Saudi Arabia in a few years (that´s the blue curve shooting up like a rocket).

To show how things work the graph also shows MY OWN GUESS. I prepared it using my scientific knowledge, a quick review of existing reserves, and a little bit of windage (windage is  what we use when we shoot moose from long range). 

My forecast (the red dots)  shows production peaking  at a bit less than 10 million barrels of oil per day (don´t be deceived by the guys who pile condensate, biofuels and all sorts of non oil liquids on top, because  they are cheating to make you feel good).

What do I think has been going on? The system has feedbacks. In this case oil supply was dropping, this in turn increased prices, and the economic environment encouraged the industry to go drill in deep water and use hydraulic fracturing in North Dakota´s Bakken. This is a feedback. And why does the production peak and drop? It´s also a feedback, in this case the oil reservoirs are plumb running empty, and I´ve assumed we don´t have another Bakken laying around to fill the gap.

Finally, the graph shows a little triangle tucked at the top. That´s Daniel Yergin´s  USA forecast for 2036. I met Daniel in the early 1990´s when we were invading the Former Soviet Union (this is the reason why I want to be President of Azerbaijan). I think Daniel blew a gasket when he predicted that figure. Or maybe the guy who passed it on didn´t realize Daniel had added condensate and other liquids to the total.

Or maybe he assumes the negative feedback I built in (the oil reservoirs empty) is offset by a positive feedback (the Bakken success encourages oil drillers to go for other Bakkens and they are found).

Strong Negative Feedback Example (from cbssports.com)

I´d like to close by pointing out that no matter which model we use, if we have a fuzzy idea of what went on in the past our predictions tend to be fuzzier. And it gets a lot worse if we don´t use all the available history to tune the equations (this is what happened with that 2001 prediction garbage I showed).

This is why we read so much about the fight between Mann and his supporters versus the M&M´s and the other guys who take shots at the hockey stick, why we read about people like Watts and Marohasy going nuts when they notice the Australians may be shifting historical temperature records, and why I´m so disappointed when I see so much work done with climate models which haven´t been calibrated using all the available history and a better understanding of feedbacks as well as the climate´s long term  cycles.

Please understand I´m not a climate wonk, so  I´m not going to get into that fight.  I´m just pointing out that,  if there are shenanigans going on with history and the way it´s used,  then they are seriously handicapped using equations and computer models to predict the future.

 Further Reading


No hay comentarios:

Publicar un comentario en la entrada