cc: "Parker David (Met Office)" <david.parker@metoffice.com>, David Easterling <David.Easterling@noaa.gov>
date: Thu Oct  7 10:46:07 2004
from: Phil Jones <p.jones@uea.ac.uk>
subject: Re: IPCC base period question
to: "Russell Vose" <Russell.Vose@noaa.gov>, Kevin Trenberth <trenbert@cgd.ucar.edu>

    Russ,
      We did have a criterion for calculating 61-90 normals (in the paper in J. Climate in
   2003 by
    Jones and Moberg). I don't think this is crucial but we need a background reference for
   you
    method if you've made any important changes since the last publication on the subject.
   With
    the HC we are updating our dataset (will go to HadCRUT3) - some new stations, but mainly
   some
    more work on outliers, checking all normals and importantly producing errors on all grid
   boxes
    as well as on the NH/SH and global series. Incorporating errors due to homogeneity checks,
    errors in normals and errors from the bucket/intake adjustments as well as the sampling
   errors
    done before.  Paper on all this in the New Year, so forget this for the first draft.
       So, answer, do what you think is reasonable and have it documented - either in a paper,
    or we can add something in the Appendix we have on many of these error issues. The paper I
    mentioned above will give a first attempt at some of the errors we've not considered
   before. It will
    be a first attempt though as many of the estimates we use are a little ad hoc. We will
   have a
    methodology to see what will happen with a range of estimates.
       Getting as many of the time series with error bars is important as Kevin says. We will
   also
    have spatial patterns as well, but these will be less crucial as we only plan to look at
    patterns over 1979-2003 (5 or 6 eventually) and also 1901-2003-5 as patterns won't change
   much
    from previous reports.
       I briefly talked to Dave about colour maps of trends (and not the dot format). We can
   look
    at this later. Need to be able to be clear where the missing areas are and the dots don't
   always
    make this obvious.
    Cheers
    Phil

   At 22:23 06/10/2004, Russell Vose wrote:

     Thanks, Kevin.
     I'm pretty sure that Phil used a few specific criteria in some of his papers (e.g., a
     minimum of 20 years of data, 4 years in each decade).  I just didn't know if these
     details had already been ironed out for IPCC (and if not, then I guess we're the ones to
     decide).
     Kevin Trenberth wrote:

     Russell, No doubt Phil will comment more.  The baseline is to establish a common period
     for anomalies and thus it anables maps of anomalies to be more coherent if the same
     procedures are used everywhere.  If there is missing data, this will clearly upset this
     consistency spatially.  In principle this should still not be a problem as long as
     uncertainties: error bars, are appropriately calculated that fully account for the
     missing data.  But while those details can perhaps be gone into in a paper it is
     difficult to do it well in IPCC.   I think there are fairly standard sorts or
     requirements for % of data required in a month, months in season, and years etc for the
     result to have some credibility and it should really be linked to error bars.  But most
     decisions have been ad hoc.
     This is where things like reanalysis can help fill spatial and temporal gaps if done
     right.  But maybe that's too ambitious here.
     All this is by way of saying I don't know the answer.  My guess is that yes it should
     have data in each decade and that it is OK to estimate a base value.  But I would let
     Phil rule on this: if he is available (may be next week).  But please do track the error
     bars.
     Kevin
     Russell Vose wrote:

     Hi guys...
     Dave Easterling thought I should drop the three of you an e-mail with a "base period"
     question.  As I understand it, the idea is to use 1961-90 as the baseline for IPCC.  But
     have any other subcriteria been discussed or otherwise set in stone?  For instance, how
     many years of data must a station have during that period?  Must it have at least some
     data in each decade?  Is it okay to estimate a normal if the station lacks sufficient
     data during the base period?
     Thoughts/feedback appreciated.

     --
     Russell S. Vose, Chief
     Climate Analysis Branch
     National Climatic Data Center
     151 Patton Avenue
     Asheville, North Carolina 28801
     Phone: (828) 271-4311
     Fax: (828) 271-4328
     E-mail: Russell.Vose@noaa.gov

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