Trawling through files from 2001 I came across this rare example of an email from Dr Jim Hansen that actually gives an insight into what GISS does with temperature data.
For background I have my page commenting on Jones et al use of Miami.
Then my page on the five degree grid cell covering much of Florida and commenting on Jones 1994 additions.
Then this page commenting on GISS data which inserts warming into rural data west of Miami.
It is indeed fascinating to ponder Dr Jim Hansen’s emails below from July 2001, I think the bottom one was first.
Consider that the IPCC draws faith in Jones et al warming trends from the point that GISS and GHCN produce similar results to Jones.
Date: Mon, 16 Jul 2001 17:54:27 -0400
To: W Hughes <email@example.com>
From: James Hansen <firstname.lastname@example.org>
Subject: Fwd: Re: Florida Rurals
Cc: email@example.com, firstname.lastname@example.org
The adjustments to these two stations seem questionable. Of course that is
true for almost all stations [see the last paragraph in section 5.2 of our
paper (still!) in press at JGR, which begins: "We reiterate a caveat that
we have discussed elsewhere --- the smoothing introduced in
homogeneity-adjusted data may make the result less appropriate than the
unadjusted data for local studies....]. The reasons for this are obvious,
adjustments based on only a small number of stations, etc., so the
expectation is only that on large area average, including global average,
the adjustment does more good than harm. On the large area average the
adjusments make the estimated temperature trend have less warming. In
these two cases it makes for slight warming in one case and larger warming
in the other. Both of these stations are "rural" in GHCN’s population
category, so they would not have been adjusted in the old GISS analysis,
but they are "dim" or "peri-urban" as seen by the satellite, so they are
adjusted in the new analysis. You could argue that it might be better to
leave them unadjusted, but it’s really not kosher to go in on a case by
case basis. Our approach is to set up an objective scheme and then let the
computer and the satellite make the decision – its not necessarily the best
way, but its objective, even though, as our caveat states, it is going to
give "bad" results at some individual stations.
If I remember right, when we looked at the satellite night lights map,
Florida was very unusual – perhaps it was that almost no "dark" sites
remained – so it may be statistics of very small numbers – I will check
with Makiko – also Florida is much harder than other places in the U.S.,
since most of it is bordered by water, so no stations there – also
meteorology is very different and small scale – lot’s of problems – but I
would not be surprised if its climate trends were quite different than the
"mainland" Southeast U.S., parts of it are in a different climate zone –
will get back to you.