cc: Sakari Uppala <ers@ecmwf.int>
date: Mon Mar  8 10:37:31 2004
from: Phil Jones <p.jones@uea.ac.uk>
subject: Re: November 1981 in CRUTEM2v, ERA-40 and NCEP
to: Adrian.Simmons@ecmwf.int

    Adrian,
        Found the problem and it is in the CRU data.  It is also in World Weather Records
    and likely in the GHCN dataset at NCDC as well. We are going to trawl through some
    files for Turkey that we have and see if we can find the correct data in them. I recall
   putting
    this data in so maybe that was wrong as well.  The data we have for Nov 81 looks like
    another month - maybe Sept or Oct.
        If we have to contact Turkey it may take a little time to get the right data. Best to
    omit that month for the time being. Also, Turkey has a lot of missing data in 1981 for
    some reason. Will try and infill that at the same time.
    Cheers
    Phil
    Adrian,
       I'll look into this. I vaguely recall some odd months during the 1980s which we
    discovered in some comparisons then. I'll have to go back to some older files
    and check with some more recent Turkish data we've received.
       During that time we were putting the CLIMAT messages in by manual means. We've since
    received the World Weather Records for Europe (RA6, so including Turkey), so any values
    should have been replaced by these numbers directly from the country.
       It may take me a few hours to get to the bottom of this.
    Cheers
    Phil

   At 12:36 06/03/2004 +0000, Adrian Simmons wrote:

     Phil
     I found a few hours to work on the reanalysis/CRU comparisons this morning, and decided
     to plot the differences between CRU and ERA-40, and between CRU and NCEP,
     month-by-month. Basically, it showed ERA-40 in a good light, assuming CRU to be truth.
     However, November 1981 over "Europe" was anomalous, as can be seen in the attached time
     series. The NCEP curve is hidden beneath the ERA-40 curve for this month: both are
     unusually different from CRUTEM2v. For ERA-40, it's the largest discrepancy in the
     series.
     I thus plotted out values for each of the grid boxes that go into the average for this
     month. The problem is confined to the south-east of the domain - strictly Asian boxes
     rather than European ones (though we have all of Turkey as a Member State!). Here are
     the anomalies relative to 1961-1990 for the three boxes with the largest discrepancies:
     Lat: 45N-40N  Lon: 30E-35E  CRUTEM2v: 7.75  ERA-40: -1.39   NCEP: -1.14
     Lat: 40N-35N  Lon: 30E-35E  CRUTEM2v: 6.85  ERA-40: -1.69   NCEP: -2.66
     Lat: 40N-35N  Lon: 35E-40E  CRUTEM2v: 4.28  ERA-40: -1.59   NCEP: -2.76
     As you can see, it is the CRUTEM2v values that are the most anomalous.
     I hope I haven't made a mistake in my processing of the CRUTEM2v data. Assuming not, I'd
     appreciate your comments on this. Are you aware of any problem with the data in this
     region?
     Best regards
     Adrian
     --
     --------------------------------------------------
     Adrian Simmons
     Head of Data Division
     European Centre for Medium-Range Weather Forecasts
     Shinfield Park, Reading, RG2 9AX, UK
     Phone: +44 118 949 9700
     Fax:   +44 118 986 9450
     --------------------------------------------------

   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
   ----------------------------------------------------------------------------
