To: Tom Wigley <email@example.com>
Subject: Re: question
Date: Mon Sep 6 11:10:47 2004
Cc: Professor David Taplin <firstname.lastname@example.org>, Ben Santer <email@example.com>
Ben should have seen the ERA-40 Report # 18. You can forward the JGR paper.
WRT 1, it is difficult to say as it depends who's produced the values. For HadCRUT2v, I
think I've convinced the HC that the globe is (NH+SH)/2. If Peter Thorne did the
then this will be the case.
There is another issue. Sometimes the trends over Jan79-Dec03 are calculated from the
300 months rather than the 25 years. Christy does this, I think.
NCDC's Globe is probably the one domain. I've been doing some work with Russ Vose at
NCDC, which he's still to write up. Most of the differences were due to how the globe
was calculated. It is more informative to also include NH and SH as well as globe in such
tables. I'll forward a plot Tom Peterson produced a week or two ago.
ERA-40 (2 )comparisons are discussed in the ERA-40 report # 18 and the JGR submitted
This also has comparisons by continent, which again are more informative. There is a plot
in that work from the full globe vs the CRU coverage. I wouldn't believe their tropics.
Antarctica is way off as well - at least where the surface data are located, so I wouldn't
have much faith in their values for the unmonitored parts.
On (3) I did some comparisons ages ago with Jim Angell's surface data from sondes. Jim's
data was just noisier and I suspect LKS would be also. I've not done anything like this
ages. The closest would be the ERA-40 comparisons, which is much more extensive than
the LKS network.
I might have a chance to do an LKS comparison if Dian sends me the co-ordinates.
Comparisons over 1958-2003 will be much more realistic, but the ERA-40/NCEP degrade
prior to the 1960s. LKS would be better here. All sonde data look odd in the late 1950s to
the early 1960s. The jump around 1976/77 has always intrigued me. It is bigger in some
regions than others - I think it gets more credence because it is large over western North
America. Kevin had a paper on this in BAMS in the late 1980s.
At 15:57 04/09/2004, Tom Wigley wrote:
On Sept. 13-17 I will be at a meeting at the Met Office to do with
a report we are writing on trends in vert temp profiles as part of the
US Climate Change Science Program (CCSP). It involves all the
usual suspects. Seven chapters, the last of which is equivalent to
a summary for policy-makers -- for which I am the lead author.
Various people are updating data sets and doing calculations of
trends, etc. Some of the surface numbers I found to be a bit
disturbing -- so I am asking for your opinion. These are trends
per decade for Jan. 1979 thru Dec. 2003 ......
SOURCE GLOBE 30S-30N
HadCRUT2v 0.169 0.127
NCDC 0.151 0.146
ERA40 0.113 0.032
LKS 0.074 0.056
(1) CRU and NCDC are consistent within the noise, but I have one
question -- how do both calculate GLOBE?
(2) ERA40 is marginally OK (relative to CRU) in GLOBE, but
the tropics is alarmingly different. (The diff here accounts for the
GLOBE difference.) Why is this? Which is better? Is this discussed
in your paper with Adrian?
(3) LKS is the surface data from the corrected LKS radiosonde data
set. The difference here must be partly due to coverage issues. But
I recall that years ago we saw a difference between surface sonde and
CRU data. Have you done a like with like comparison (i.e., selecting
the LKS sonde sites and extracting the corresp CRU (and NCDC, and
ERA40 -- and (if possible) NCEP) data? This seems to be a pretty
basic sanity check on the sonde data -- so, if you have not done this
already, could you do it for me please?
I think there is a nice little GRL paper here. For the CCSP we are also
giving trends, etc. over 1958-2003. So the real need is for a full time
series comparison over this period -- i.e., not just trends. In other
words, what I would like you to produce is the monthly time series
for the various data sets for the LKS coverage. If you don't know
the LKS site locations, I can get these for you.
Re going back to 1958, the sonde trop data have a well known (but
not well explained) problem over roughly 1958 to 1964/5. I am curious
as to whether this shows up in the LKS surface record. I am also
curious about the apparent 1976 jump -- some people have made a
lot of noise about this, but I don't see it as a major item in the global
surface data. So the Q here is, is is apparent in the restricted coverage
of the sonde data?
I hope you can help. I am leaving here on Sept 7 to spend a few days
with a friend of mine in Plymouth -- you could contact me thru him (I
am copying this to him so you can see his email).
Prof. Phil Jones
Climatic Research Unit Telephone +44 (0) 1603 592090
School of Environmental Sciences Fax +44 (0) 1603 507784
University of East Anglia
Norwich Email firstname.lastname@example.org