Thursday, January 19, 2012

2085.txt

cc: t.mitchellatXYZxyz.ac.uk
date: Tue May 28 09:20:43 2002
from: Mike Hulme <m.hulmeatXYZxyz.ac.uk>
subject: Re: Fwd: comments for Helsinki
to: Timothy Carter <tim.carteratXYZxyzaristo.fi>

Tim,
I had forgotten that you were doing the supra-regional pattern-scalings for TGCIA - but now
I remember! As well as the four seasons, ANN would be useful also. Of course what Tim has
done here is not pattern-scaling - just scatter plots.
Of course the country versus supra-region issue is of concern (as I have mentioned to Tim
M), but I take the pragmatic line that many users (e.g. AIACC) find some guidance better
than none. And of course CIESIN now are producing country-level pop and GDP using perhaps
even more heroic assumptions that we have to make in manipulating GCM data. These matters
need discussion and until this has been had none of Tim's results will go on the DDC -
although they are available through his personal web site.
I agree about effort and resources which again is one of my points (and an additional
reason why I saw my time as up) - the work needed to be put into the DDC to make it
increasingly useful is more than can be covered by goodwill or by mere dribs and drabs of
money.
Anyway, the Helsinki discussions will be important with Richard Moss taking over from
Martin.
Good luck!
Mike
At 16:01 27/05/02 +0300, you wrote:

Mike,
Thanks for the comments. I am sorry, though not too surprised, to hear that you are
stepping down both from the DDC and the TGCIA - pity!
I knew that you were preparing country-level scatter plots, but I wonder how these mesh
with the pattern scaling work we are currently undertaking for the Task Group. We are
still using the 32 regions we used previously, and for all four seasons. In fact, the
pattern scaling component will probably be rather minor (perhaps unnecessary for 2020s
and 2050s as inter-model differences dominate; 2080s needed for A1FI), as the direct
model outputs already offer quite a large spread of results. That work is ongoing, but I
wonder how your set of country plots can be harmonised with the regional scatters. They
are likely to be fired on by some of the GCMers for being too grid box specific (in
small countries). The data Richard Jones made available from the TAR models would also
be useful for us to use at regional scale, although they are only for DJF and JJA, since
not all of these models are yet available on the DDC.
I hope that my colleagues at the Met. Institute don't feel that their comparable work to
construct scatter plots for Finland in the FINSKEN project has been too much in vain,
given that you have presented Finnish plots using Richard's dataset! Nevertheless, it
will be a useful exercise to compare the two sets. One thing you and Tim might like to
look into is the performance of these models in simulating present-day climate. There
are some very strange and clearly unrealistic results over Finland from some of the
models for the 1961-1990 period (the baseline we generally use) that have been revealed
in our analysis, even without needing to compare directly with observed climate.
Moreover, we have found some examples (looking at the three time slices independently)
where scaling appears not to be well justified. These are features that really argue for
some diagnostic information about the model performance being added to the DDC,
especially if we are to present regional/country information. I see you have highlighted
this as one of your discusion points.
Guidance is perhaps the major issue here, and I admit that I have not been sufficiently
active in following this up (too many other things to do). Someone really needs to take
this issue and run with it, full time, to come up with a useful product, preferably
online.
Perhaps you are wise to pull out after all ......... and England will probably get
another hiding to boot!
Best regards,
Tim

No comments:

Post a Comment