On 08/12/2015 10:43, Chris Hogg wrote:

One can argue that to select any section of the available data is
cherry picking. For example, it's just as valid to argue that the
period from 2000 to present is flat as it is to argue that the period
from 1970 - 2000 showed significant warming.

No it isn't and you more or less got the reason why in what you said
later. If there is a lot of noise relative to the underlying trend, any
conclusions about trends or statistical significance will be more
unreliable. If you have a short period contained within a longer period
what you can do is to analyse whether that shorter period is consistent
with the overall trend and whether there is any evidence that the trend
might have changed. It is not sufficient to eyeball the graph and say
that that bit looks a bit flat. Even fairly simple techniques based on
linear regression show that the data post 1998 is well within the range
of variability around the same sort of continued trend, and more
involved change point analyses have come to a similar conclusion.

On the other hand, my understanding is that similar sorts of analysis do
pick out some changes in trend over the 20th century which you yourself
highlighted...

If cherry picking data is one's game, how about the warming period in
the thirty years from 1910 - 1940, when temperatures rose by about
0.45 deg.C. comparable to the thirty year period 1970 - 2000 which
exhibited a similar temperature rise. The former clearly wasn't due to
anthropogenic CO2, but doesn't seem to attract much attention. Or what
about the period of cooling from say 1958 to say 1976, shown here
http://tinyurl.com/qfkc78c taken from the same reference as above.

There's a logical fallacy in the it is possible to have similar
effects arising from different causes. For example, solar activity was
rising in the first part of the 20th century and aerosols peaked in the
decades following the war. On the other hand, solar activity is
relatively low now, aerosols have probably been relatively high in
recent years, there's no reason to expect temperatures to be increasing
because of Milankovitch cycles etc, and yet temperatures are on an
upward trend.

At the end of the day, you can throw your hands up and say it's all
random fluctuations or you can look for physically plausible
explanations. The skeptic in me says that you don't get random
fluctuations on the scale we are talking about and I'm also persuaded by
the fact that this was all predicted before it actually started being
measurable. It is possible to dig out the literature and confirm this
with ones own eyes, and I'm not sure that even TNP would argue that the
global conspiracy began in the 60s. Or maybe he would.


In fact none of the variations are actually statistically significant,
as has been discussed here http://tinyurl.com/pvfgn5d and here
http://www.informath.org/AR5stat.pdf (thanks to TNP for those). I
don't pretend to understand the detailed statistical arguments, but
the gist is clear. It doesn't mean that the fluctuations didn't
happen: they clearly did, but in the grand scheme of things they're
not something to get too worked up about.


The problem with those references is that they are nothing *but*
statistical arguments. They boil down to data fitting without any physics.