« T'internet t'save us all | Main | Making Education Pay »

MIT The Sky is Falling

Global warming of 7C 'could kill billions this century' - Telegraph

The study, carried out in unprecedented detail, projected that without "rapid and massive action" temperatures worldwide will increase by as much as 7.4C (13.3F) by 2100, from levels seen in 2000.
Previous estimates have concluded that the likely increase this century would probably be 2.4C (4.3F).
The study uses the MIT Integrated Global Systems Model, a detailed computer simulation of global economic activity and climate processes that has been developed and refined by the Joint Program on the Science and Policy of Global Change since the early 1990s. The new research involved 400 runs of the model with each run using slight variations in input parameters, selected so that each run has about an equal probability of being correct based on present observations and knowledge....
Prinn says. "The least-cost option to lower the risk is to start now and steadily transform the global energy system over the coming decades to low or zero greenhouse gas-emitting technologies."

The first line of their home page says - The question is no longer whether global warming is upon us … but how we can rise to its challenge.. That will be the global warming of the last ten years or the global warming of the last 400? The first doesn't exist and the second is a gentle steady natural process so I guess not. I expect it means the global warming from when you last held hands with a real girl in the late 1990s...

"Least-cost option"? Destroy civilisation because of a computer read out based on what they admit are " so many uncertainties, especially with regard to what human beings will choose to do and how large the climate response will be, "we don't pretend we can do it accurately. Instead, we do these 400 runs and look at the spread of the odds." ? I don't think so Prof.


400 runs given each an equal chance of being correct means each has a 0,25% chance of being right. I am so worried! Not by the study, but by the fact this crap gets into print.

I've done my share of computer modelling and it is ill suited to the problem of long term climate projection.

A problem like modelling an aircraft wing is tractable because it is constrained and testable within those constraints: if you divide the wing size by 10 and run the same model, you'll almost certainly get bad results because you're trying to run the model outside of its tested range.

The climate is neither constrained nor testable, so you won't get any useful predictions out of climate models.

When one of those runs begins with known conditions from, say, 1950 or 1960 or 1970, and accurately predicts the experience of the intervening years, I'll back them. Until then, it may or may not be an interesting problem in computer science, but it has no meaning otherwise.

Post a comment