date: Tue Nov 10 09:18:48 2009
from: Phil Jones <p.jones@uea.ac.uk>
subject: Re: LvsO
to: Tom Wigley <wigley@ucar.edu>

    Tom,
       The separate land and ocean datasets are merged around coasts and islands (at least for
   HadCRUT3). This means that you can't get back to the separate versions from the combined.
   If you look at the grids for a particular month you'll see many points the same in the
   separate land and ocean datasets, but weighted averages around the coasts and islands.
      We've been doing this since the early 1990s. We have altered how we do it. The next
   version will go back to a simpler weighting method than used in Borhan et al. (2006). Some
   grid boxes are in both.
    Differences will likely vary from year to year and within years due to sea ice.
     As for Anders Moberg's paper, I've no idea what NHT series he used. I'm presuming you
   mean his reconstruction as opposed to the instrumental series against which he calibrated
   the proxies. Splitting the variance into high and low probably has serious problems. All
   the high freq (<80 years) comes from trees with all the low freq coming from the other
   essentially uncalibrated proxies.
      Anders is OK, but I don't think he'd want to revisit this series. He tried to consider
   the implications of variance splitting in the attached paper. I recall it not being very
   convincing.
    Cheers
    Phil
   At 21:25 09/11/2009, Tom Wigley wrote:

     Hi Phil,
     Re land versus ocean, there is an interesting problem
     here, which may apply to all data sets. If I weight
     L and O "correctly" (not straightforward) then the
     weighted sum of L (CRUTEM3v) and O (HADSST2), as a
     global-mean time series will not be the same as the
     global mean direct from HadCRUT3v. This is because
     (I think) there are some gridpoints that appear in both
     the L and O data.
     Is this correct?
     I was rather surprised at how much my two versions of the
     L and O averages differed. However they were merged, I'm
     sure it could be done better, so as to minimize any
     inconsistencies like the one noted above.
     -----
     On another point, what do you think of the Moberg (2005)
     series for NHT? I know you have worked with him, so he
     can't be all bad -- but this series is not only rubbish
     in terms of its variability, but it has been used by some
     arch skeptics to support their own silly ideas.
     Perhaps there is a chance that Moberg could re-do his
     series better?
     Tom.

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