I'm no statistician, but I believe that mean temperatures will conform to a
"normal" distribution.

You are simply saying (I think, and without quantifying it)) that the
standard deviation of a set of "chaotic" samples is greater than that of
"random" samples. But "standard deviation" should be meaningless if the
distribution is random.
....
That this is not
the case is clear from looking at any set of monthly mean temperatures.

Martin
Not sure what anybody is trying to get at here but current thinking is that temperature records shld
exhibit power law scaling typical of non linear systems.For an apparent contradiction to this in the
CET-


Scaling of Central England Temperature Fluctuations?
Joanna Syrokaf1 and Ralf Toumif2
Abstract
Central England temperature fluctuations are found to be monoscaling with long-range dependence.
Monoscaling can be explained in terms of the dominance of Gaussian temperature advection.
Simulations of the UK Meteorological Office Hadley Centre general circulation model do not capture
many of these features.
Atmospheric Science Letters
Volume 2, Issues 1-4 , June 2001, Pages 143-154
http://www.sciencedirect.com/science/journal/1530261X

also see-
http://www.copernicus.org/EGS/egsga/...ts/aai0997.pdf

--
regards,
david
(add 17 to waghorne to reply)