Friday, March 07, 2008


Predicting the Weather is Hard

Glenn Reynolds at Instapundit often says that he will believe global warming is a crisis when those pushing the theory begin to act like it is a crisis. I'm not sure that would do it for me. If the best scientific minds can't get the short term predictions right, like the weather tomorrow, or even next year, why in the world would we believe they know what the weather will be like in a 100 years. The real truth is that they don't. They have no freakin' idea and it is the worst form of hubris for them to think they can accurately predict the future weather over such a temporal distance.

Here is their prediction last October for what this finally ending Winter of 2007-2008 would be like. The NOAA said it would be warm, (December through February is likely to be another milder-than-average winter for much of the country) but it's actually been the coldest winter in decades.

The U.S. National Climatic Data Center (NCDC) reported that many American cities and towns suffered record cold temperatures in January and early February. According to the NCDC, the average temperature in January "was -0.3 F cooler than the 1901-2000 (20th century) average."
Not the stuff for building confidence in their ability to predict.

(h/t Boortz via Mark Dunn) Boortz has a lot bad to say today about the Colorado public school system too.

UPDATE: Prince Charles says it is madness to doubt that the Earth is warming primarily due to man made CO2. Oh, OK. If an intellectual powerhouse like the passed-over Prince says I'm being stupid, then that's good enough for me. I'm sorry for going 'crazy' on you. That really cold temperature you feel right now in most of the Northern Hemisphere is probably just the product of your delusional skepticism. It's clearly getting warmer and just as clearly we powerful humans are the blame. Carry on.

UPDATE 2: Just in case you didn't know I was being sarcastic above, I was. I have checked the Cryosphere Today website. The sea ice in the North is not melting, despite the misinformation from the Prince. Half of the 14 distinct areas have more ice than normal, three are just at normal and four are down--Greenland Sea, Hudson Bay, Sea of Okhotsk, and Barants Sea. In the South, the sea ice is no longer melting and is nearly 100,000 square Km above normal. There is, therefore, in the whole world, more sea ice than normal. We'll see what happens in the North this Summer. I'll go out on a limb and say it will not melt below 75,000 square Km below normal. Any Warmies care to make a gentleman's wager? Tone?

UPDATE 3: Here is a graph of sea level rise for the past 9 thousand years. Wow, it's really been shooting up over the past 2,000, hasn't it? (More sarcasm). Despite the no longer a boy Prince's pronouncement about recent startling sea level rises (about 6 inches in 130 years), the real rise in sea levels was happening 8 thousand years ago, well before any man made CO2, other than a few camp fires and of course the out breath of a few hundred thousand cave men. This view of course gives no credence to the alternative reality 'histories' of Atlantis and Mu. I wonder if Prince Charles has a position on those myths as well?


While, as I've noted here before, I think the dominant temperature driver is solar output, you're repeating an statement that reduces the strength of your argument:

"If the best scientific minds can't get the short term predictions right, like the weather tomorrow, or even next year, why in the world would we believe they know what the weather will be like in a 100 years."

If the best mathematical minds can't get the short term predictions right, like what the winning number on the next spin of the roulette wheel will be, why in the world would anyone think that casinos would make money?

Local temperature is extraordinarily noisy and global average temperature (using any method of determining it) is quite noisy as well. Being incapable of predicting the noise in the signal is pretty much what makes it "noise" and not "signal".
Goood points, Doug, but I think you do a little intellectual sleight of hand about the Casinos. Good about the noise though. But if the Warmies are predicting a global average temperature rise by 2100 of 3 degrees C, are they not telling us what the weather will be, on average, then? It's not like they are predicting what sellers of carbon indulgences will net by then, they are still predicting weather.
They're predicting mean weather, not weather. Similarly, casino games are designed to predict mean payout (and ensure it's below 1:1). Even a long run of marginal losses doesn't say much about whether the game is flawed. (But the longer the run, the more likely there's a systematic problem, of course.)

There are several problems with climate prediction (and this is particularly so for predictions of climate change, though some people are reluctant to realize it):

1) The signal to noise ratio is way below one. That is, hourly, daily, weekly, monthly, and yearly random variations are much greater than the supposed change in the baseline.

2) Systematic cyclical variations (morning to night, season to season, beginning to end of sunspot cycle, whatever) are hugely greater than any change in the baseline.

3) The length of the reliable baseline is very short relative to the noise in the data.

That last one is particularly important. In a casino, there are probably 100's of thousands of individual events every day, each of which is completely measurable. Any change in probability will show up very quickly, even if the change is very small. For planetary temperature, you only get one year-event each year (or pieces of a half-dozen overlapping sunspot cycles every decade).

People have been trying to extend the baseline through a variety of innovative methods (tree ring studies, ice cores, whatever) but the data from those sources is quite unreliable. And it's deeply unclear just what might have been causing any of those historical changes anyway, which makes deriving the cause of any current change even more fraught.

It is precisely this issue that has caused most of the problems in prediction. The conundrum is that there's really nothing much that you can do about this. And the result of that is that you are trying to make important decisions on weak information.
Here is where the casino analogy is flawed. The mathematical certainty of the probability of given events over time--the casino's, not the player's life--demonstrate that flawless (no cheating, manipulation, etc.) will provide the house an "edge" that is both guaranteed and insurmountable by even random variations. No one suggests that the odds on a roulette table will change due to the lighting, the casino's carpeting, or any other external factor. Should a roulette wheel be "biased" or a evidence of card counting observed, the casino will replace the wheel and the player accused of counting will be tossed out. Even slot machines must meet state requirements, or be replaced.

On the other hand, meteorologists and climatologists both insist that nature is immensely complex, and that even the slightest variation will cause the best computer models to be flawed, and predictions to become incorrect. If there are 20 variables, for instance, that govern a weather forecast that is obviously not entirely accurate--based as much upon past events as it is on current observable conditions, then how can an even more complex system of climatological variables be any more easily predicted.

A roulette wheel in Las Vegas or Macao will (should mathematically) behave in the same way over time. It is clear that the world's climate varies greatly over time, without the single addition of any human-derived element.

Comparing the two is simply obfuscation.
Post a Comment

<< Home

This page is powered by Blogger. Isn't yours?