cc: l.alexander@bom.gov.au, Albert.Klein.Tank@knmi.nl, Jay Lawrimore <Jay.Lawrimore@noaa.gov>, Byron Gleason <Byron.Gleason@noaa.gov>, david.easterling@noaa.gov, aiguo dai <adai@ucar.edu>, David Parker <david.parker@metoffice.gov.uk>
date: Mon Jul 18 09:38:25 2005
from: Phil Jones <p.jones@uea.ac.uk>
subject: Re: Fwd: RE: [Fwd: Re: [Fwd: Re: Some more thoughts on DTR map for
to: Kevin Trenberth <trenbert@ucar.edu>

    Dear All,
       I'm away after today until Friday. I've arranged for Aiguo to get the latest
    CRU data. Also, I've sent some of the plots on to Karl Taylor (at PCMDI).
    Karl is writing for Ch 8 on model evaluation and is checking a plot of
    zonal mean abs DTR against the models. He's using the CRU data as
    the obs absolute. This isn't that relevant for the trends, but we need to be aware
    that most models underestimate abs DTR by between 2-4 K.
       I've suggested he sticks with CRU data as he has it. It would be useful though
    to get abs DTR for comparison from NCDC at some point.
    Cheers
    Phil
   At 16:15 15/07/2005, Kevin Trenberth wrote:

     Hi all and many thanks to Lisa for thie maps.  Some points I would make
     1) The trends are not significant locally but when they are the same sign over large
     areas, one gains significance.  It means that small differences, like Aiguo using 1979
     to 2004 and  Lisa 1979 to 2003 can make a difference, and 1950 vs 1951 for the start.
     The end points matter a lot with trends.  It also suggests that this ought not to be
     done with fine grain structure but rather with a much coarser grid.
     2) The arguments about significance apply for all fields not just DTR.  This has long
     been an issue with the MSU record, for instance: all the differences in dispute are in
     fact not significant mostly.  Since we decided to go with 1979-2004, I think we should
     try to stick with that and see if we can get a coarse grain product.
     3) There are nonetheless major differences with the Vose product in places like Spain
     and South America that clearly require a lot more work from NCDC to sort out.  Aiguo has
     the advantage of a whole lot more data from all the synops, but the disadvantage of not
     having the fine quality control.
     4) Lisa's map seems to have colors in it that are not in the color bar.  (Or vice versa:
     the color bar is weird)  Most places the agreement is not that bad with Aiguo, Australia
     being an exception.   Phil noted that Aiguo has been using an older version of the CRU
     data.
     5) It may be that we need to look at some time series to see what years are causing the
     trends to be so different in a few places.  e.g. Spain, Argentina,  some spots in the
     US?
     Got to run
     Kevin
     Phil Jones wrote:

      Lisa et al,
         Thanks. Forwarding to the others at NCDC involved in all the discussions.
      I think the main conclusion to draw from all this is that 1979-2004 is too short
      a period.
          If Russ produces a map for 1951-2004 then hopefully they will all look much
      more similar. We should likely go with one of these plots. We probably need
      to consider if we want a trend map. I would say yes, as if we get the right one, it
      will show that DTR is decreasing as an average, but definitely not in many locations.
          There are certainly lots of potential problem areas as Lisa alludes to, as does
      the detail from Aiguo.
      Cheers
      Phil

     Reply-To: [1]<l.alexander@bom.gov.au>
     From: "Lisa Alexander" [2]<l.alexander@bom.gov.au>
     To: "'Kevin Trenberth'" [3]<trenbert@ucar.edu>,
             "'Phil Jones'" [4]<p.jones@uea.ac.uk>,
             "'Klein Tank, Albert'" [5]<Albert.Klein.Tank@knmi.nl>
     Subject: RE: [Fwd: Re: [Fwd: Re: Some more thoughts on DTR map for 1979-2004]]
     Date: Fri, 15 Jul 2005 20:07:11 +1000
     Organization: BMRC, Bureau of Meteorology - Australia
     X-Mailer: Microsoft Outlook, Build 10.0.2627
     Hi all,

     I have plotted up the 1979-2003 DTR trends using the global extremes data.  Ive tried to
     use a similar colour scheme and range to Aiguo since he also has trends for the longer
     period.  Things to note:-

     1.  I have only used quality controlled stations which have at least 40 years of data
     between 1951-2003.  This is because otherwise I would have had to regrid all of the data
     which would take some more time to complete.  The spatial coverage would be better if I
     say regridded using stations that had at least 20 years of data between 1979 and 2003
     and this might be something worth considering.

     2.  Figure caption:-  Observed trends per decade for 1979-2003 for the diurnal
     temperature range index.  Trends were calculated for grid boxes which had at least 20
     years of data during this period and ended no later than 1999.  Black lines enclose
     regions where trends are significant at the 5% level using the method of Wang and Swail,
     [2001].

     3. I have used the HadCM3 grid size (2.5 x 3.75).

     4. There are very few areas showing significant change.

     We can safely say that there is not general agreement between the 3 results!  I agree
     well with Aiguo in some regions and with Russ in others.  For comparison I asked Dean
     Collins to plot DTR trends for Australia for the two periods (attached).  I would say
     that Russ and I agree better over Australia than Aiguo.  Since Dean and I source the
     same daily data our results should be similar but there are some differences in detail
     which would need to be checked out.  Aiguo and I agree better in terms of the sign of
     the trend over South America but not in magnitude.

     I have not read Russs or Aiguos papers so I cant comment on their methods.  Potential
     areas of discrepancy are:

    1. Are we all using exactly the same definition of DTR?
    2. Data sources.
    3. Different gridding methodologies.
    4. Different trend calculation methodologies and missing data tolerance.


     One might suggest that the sample size is too small to accurately calculate trends over
     25 years.  However it is quite worrying that the sign of the trend can be so different
     between the methods.  When I calculated the correlation decay distances for the paper,
     DTR was one of the less coherent indices which may be a factor.

     For some reason Im having trouble getting the percentile maps and timeseries diagram
     exactly how you want them.  Perhaps its because its Friday.  Im attaching what I have so
     far.  Y-axis now represents the anomalies in days.  Perhaps Ill have better luck on
     Monday.

     Regards.

     Lisa.

     -

     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    [6]p.jones@uea.ac.uk
     NR4 7TJ
     UK
     ----------------------------------------------------------------------------


--
****************
Kevin E. Trenberth                              e-mail: [7]trenbert@ucar.edu
Climate Analysis Section, NCAR                  [8]www.cgd.ucar.edu/cas/
P. O. Box 3000,                                 (303) 497 1318
Boulder, CO 80307                               (303) 497 1333 (fax)

Street address: 1850 Table Mesa Drive, Boulder, CO  80303

   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.jones@uea.ac.uk
   NR4 7TJ
   UK
   ----------------------------------------------------------------------------

