date: Mon Jul 18 09:01:20 2005
from: Phil Jones <p.jonesatXYZxyz.ac.uk>
subject: Re: need daily max/min or dtr climatology
to: Karl Taylor <taylor13atXYZxyzl.gov>
The earlier email should set you up for collecting the new CRU TS 2.1 dataset.
Your email about DTR is quite timely considering all the email traffic we've
had on the issue over the past week. I'll try to briefly summarise this and send
you some figures. Most relate to trends over the periods 1950/1-2003/4 or
1979-2003/4. Apologies for resending your figures, but together with the others
we have, they are useful background.
In answer to 2) there are 3 versions of DTR that we've been looking at:
- one from NCDC, Russ Vose, a trend map so far for 1979-2004
- various plots from Aiguo Dai at NCAR
- plots for 1950-2003 and 1979-2003 from Lisa Alexander (who is an HC
employee but works at BoM in Melbourne at the moment)
After your three the png file is from NCDC (Russ Vose, who's currently away).
The next is from Lisa for 1951-2003
The next two are from NCDC and are trends over the USA
and the final two are from Aiguo Dai and include some other fields like cloud
All this may be more than you want, but these are the three versions we are
potentially working with. If you look at the trends for the approximately the same
periods, you will see the problems we have. Trends are often opposite for some
regions (southern South America, Spain). There are some other plots as well, but
these are enough for the moment.
We will likely go with a trend map for 1951-2004, as these are closer. The differences
for 1979-2004 reflect problems, but also are due to the period being too short.
So, this answers 2) and also 4). Lisa is HC, so her version is essentially what the
HC could produce.
NCDC's version uses thousands of stations (~5000).
Lisa's uses stations that have been collected through workshops, so
QC is likely better.
Aiguo's is CRU up to a point then has synop data for more recent years.
Ch 3 needs to resolve these trend differences. We may not fully succeed in time for the
FOD, but we need to go with one of these.
As for 1), I'd suggest you use CRU TS 2.1. This will be globally complete. I would
suggest you only go with the period 1951-2004 or a part thereof (i.e 1961-90 if you
want). CRU is globally complete (except for Antarctica), so omit 60-90S (you seem
to have done this). Also models should be land only. The best zones for the obs are
70-30N and 20-40S. Tropics less good, but I reckon the obs are not that much
Greenland isn't good in CRU. Is there a possibility of plotting a model-obs field
and seeing that? Do you adjust the models for elevation? DTR tends to be higher
with elevation. DTR also has a marked seasonal cycle. Can you see if the models
are closer in DJF than JJA, for example?
Hope these few ideas help and you can get access to the data. Kevin may
have some other ideas.
PS I'm away after today and not back in CRU till Friday.
At 00:11 17/07/2005, Karl Taylor wrote:
For the model evaluation chapter of the IPCC AR4 we need the climatological monthly
means of daily Tmax-Tmin. Celine Bonfils, who is working with me, obtained what we
needed (at .5x.5 degree resolution) from your web site and plotted the zonal mean (over
land areas only). We also took the difference between the models' diurnal temperature
range (DTR) and the observed. I've attached a few figures showing samples of the
results for annual means of DTR.
I have a few questions:
1) Do you think the observations are reliable enough to conclude that the model's
largely underestimate DTR?
2) Are there alternative obs. data sets available?
3) Who should I contact at the Hadley Centre to get their HadTX and HadTN data?
4) How does your DTR climatology compare to the one at NCDC, analyzed by Tom Karl and
I'm trying to get my section of chpt. 8 finished, so any quick responses would be much
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 p.jonesatXYZxyz.ac.uk