date: Fri, 23 Dec 2005 16:12:23 +0000
from: Gerard van der Schrier <schrier@knmi.nl>
subject: first version of ms.
to: Tim Osborn <t.osborn@uea.ac.uk>, Keith Briffa <k.briffa@uea.ac.uk>

<x-flowed>
Tim & Keith,

Attached is a first version of a ms. about the optimal fingerprinting 
technique, and an application to relate changes in sea-surface height 
(SSH) to changes in oceanic meridonal heat transport (MHT).

There are two major differences with an earlier version we discussed 
some time ago. The estimates of trends in MHT from trends in SSH is 
beefed up some more. The correlation between observed and estimated 
changes in MHT is now 0.71, which is not too bad. The 95% confidence 
levels decreased to 1.39, which is still pretty large (the trends in MHT 
are normalized, so the confidence limits are 1.39 sd.......). On the 
other hand, this would relate to a trend of 0.039 PW/yr. Bryden et al. 
recently found a change over the 1990s of 0.33-0.5 PW/yr.

The other thing is that my estimate of a change in MHT over the 1990s, 
based on altimetry data, is now negligable. This is not necessarily a 
bad thing. Knight et al. (2005) recently estimated an *increase* in MHT, 
Bryden et al estimated a *decrease* and Carl Wunsch has a submitted 
paper in which he estimates the change in MHT to be negligable too.

The ms. still feels a bit 'thin'. I've tried two approaches to extent 
this analysis. One was Tim's suggestion to look at trends of 5, 20 and 
40 years too. You see that the skill of the method increases somewhat 
when the length of the trend is increased to 20 years, to decrease again 
for 40 years. The improvement in skill, and the associated decrease of 
the size of the confidence limits, is not very impressive.

The other way was to use multiple patterns to estimate trends in MHT. 
For some reason, this did not yield better estimates of MHT compared to 
the univariate approach. This is somewhat counter-intuative though. I 
don't know why the skill does not improve.

Cheers, Gerard

-- 
----------------------------------------------------------
Gerard van der Schrier
Royal Netherlands Meteorological Institute (KNMI)
dept. KS/VO
PO Box 201
3730 AE De Bilt
The Netherlands
schrier@knmi.nl
----------------------------------------------------------


</x-flowed>

Attachment Converted: "c:\eudora\attach\artv13.pdf"
