- China View: Ten thousand EU workers protest in Brussels against the "climate" plans to cripple Europe's steel industry
- PhysOrg: German carmakers denounce the EU "compromise" on CO2 emissions
- World Climate Report: Will one of the 10,000 participants of the Poznan climate conference notice that there hasn't been any warming for 10 years?
- Deutsche Welle: Václav Klaus, a leading AGW "doubter", will become more visible during the Czech EU presidency
- Investor's Business Daily: Cooling down: people will realize that there exist serious issues, too
- Climate Progress: Maybe, Obama is not the ultimate green messiah, after all
- WSJ blog: Analysts are scrabling to dampen already low expectations in the Poznan summit
- Orange Punch: The Terminator and friends have edited their warming books to make his heroic struggle to terminate climate change look much cheaper
As a part of the refreshing of my basic Mathematica skills, I've created this simple self-explanatory Mathematica notebook explaining why Michael Mann gets the hockey stick even from random data - Brownian noise.
In order to make the effect really nice and obvious, I first prepare 5,000 random Brownian temperature proxies, obtained by a resummation of random independent numbers between -0.5 and +0.5 over time. Click the picture below to zoom in.
In the second step, I create "instrumental" 20th century data. To simplify a bit and extract the qualitative message of this whole homework, I assume that the temperature was increasing linearly. At this moment, I am ready to compute the correlation coefficients of each of the 5,000 proxies with the "instrumental" temperature data. The average temperature of the proxies, weighted by the correlation coefficients - telling you how "good" the proxy is according to Michael Mann - is drawn below: click to zoom in.
What we're doing is completely obvious. We're choosing the "good" and "important" proxies by their having an unusual 20th century trend, without paying any attention to their behavior in the previous centuries. It follows that the behavior of the random proxies before 1900 averages to zero while the selection criterion for the 20th century survives. The result is inevitably a hockey stick.
Let me emphasize that we have entered random data from a virtual world where the temperature in the 20th century had the same variability as the temperature in the previous centuries. Nevertheless, the Mannian method has transformed the random data into a hockey stick. That's the simplest way, due to M&M, to prove that Mann's methodology is complete rubbish.
If you're worried about the linear 20th century, it doesn't really matter because the 20th century trend was pretty similar to the linear function and with the exactly measured 20th century temperatures, we would obtain (and Mann obtained) qualitatively similar data. Still, the appearance of the hockey-like graph clearly doesn't mean that the 20th century was any special: in our world, we were very careful to guarantee that the 20th century was generic and ordinary.
Using Mann's method, you get the hockey stick if the data are random and unrelated to the temperature. But even if the data are correlated with the temperature, it is easy to see that the resulting hockey stick doesn't imply that the real temperatures looked like a hockey stick. In fact, the temperatures could have been linearly increasing from the year 1000 to 2000 - imagine that you add this "real" linear function to all the proxies. The result, using Mann's method, would still have the hockey stick, superimposed onto a negligible linear trend. You can play with the program yourself. If you add any "real temperature profile" to the Brownian motion, Mann's method still gives you the "real temperature profile" plus (superimposed on) the hockey stick (that comes from the random contribution to the proxies, more precisely from the bad proxies that were overrepresented because they agreed with a warming trend "by chance").
To put it simply, Mann misinterprets the 20th century warming as a proof that the warming was unprecedented, uses this misinterpretation to select proxies, and "surprisingly" derives his assumptions back. If you have Mathematica, you may download the notebook here: