from: "Nigel Arnell" <n.w.arnellatXYZxyzding.ac.uk>
subject: RE: QUEST-GSI timings
to: "Tim Osborn" <t.osbornatXYZxyz.ac.uk>
Thanks - that's very helpful. From the data you sent me yesterday, I was
leaning towards using individual model tunings rather than the mean.
I'll adjust the experimental design note accordingly.
The outlier is MIROC...
Professor Nigel Arnell
Walker Institute for Climate System Research
University of Reading
From: Tim Osborn [mailto:email@example.com]
Sent: 15 July 2008 09:31
To: Nigel Arnell
Subject: RE: QUEST-GSI timings
yes, I agree that the climate uncertainty is bigger than the carbon
cycle uncertainty, for determining global-mean T. Especially at mid
century, but even at 2100. So, taking multi-model-mean global T will
miss an important part of the uncertainty range.
Reasons not to use specific model (or at least MAGICC approximation
to specific model) global T? The only one is if the particular
selection of GCMs is biased towards a particular part of the
uncertainty range, or even if some GCM(s) have sensitivity that lies
outside the IPCC likely range. The attached table indicates the
climate sensitivities of the models being considered. Even the
apparent "outlier" at the top in your temperature plot (which comes
from MIROC hires I think) has an estimated equilibrium sensitivity of
4.3 K, i.e. within the IPCC likely range. So, this doesn't seem to
be a reason.
So, seems best to use individual model glob T, capturing the climate
uncertainty, but ignoring the much smaller carbon cycle uncertainty.
For QUMP runs, your suggestion of using the HadCM3-tuned version of
MAGICC global temperature is best. All differences between runs
would be, as you say, due to internal physics generating different
spatial responses. However, the effect of internal physics causing
different global temperature responses would not be captured, since
they would all follow the standard HadCM3 behaviour.
The only other idea that comes to mind is that, if we know the
equilibrium climate sensitivity of each QUMP run, we could re-run
MAGICC with all standard HadCM3 parameters except for equ. clim.
sens., which could be varied from one QUMP run to the next. This is
more complicated and thus more work, and also I'm not sure what we do
if some of the QUMP runs have equ. clim. sens. that lies outside the
IPCC likely range.
Hopefully the attached table is useful,
At 16:32 14/07/2008, you wrote:
>Thanks for the global temperature changes.
>I've been looking at the relative uncertainty due to carbon cycle
>feedback and model parameter set, and the latter source of uncertainty
>is of course dominant (see attached figure for A1b). By the 2050s,
>scaling by high, rather than medium, carbon cycle feedback just alters
>the climate changes b less than 5% (e.g. a 20% increase in rainfall
>would become a 21% increase in rainfall); the differences would likely
>By scaling by low/medium and high carbon cycle feedback, and averaging
>across all 19 models we are missing much of the uncertainty in climate
>change at a given time horizon. We would get a much greater range by
>simply scaling each model pattern by its own change in global
>temperature delta T (just assuming medium carbon cycle feedback). I
>that some of the fits are not very good, and that we don't have tuned
>parameters for the QUMP runs, but are there other good reasons why we
>don't rescale by the delta T appropriate for each model? We could
>rescale QUMP by the HadCM3-tuned parameters, and all the difference
>between runs would be due to internal physics.
>Professor Nigel Arnell
>Walker Institute for Climate System Research
>University of Reading
>From: Tim Osborn [mailto:firstname.lastname@example.org]
>Sent: 14 July 2008 14:43
>To: Nigel Arnell
>Subject: Re: QUEST-GSI timings
>At 14:35 14/07/2008, you wrote:
> >Ps - could you email the MAGICC spreadsheets with change in global
> >average temperature for each model pattern?
>I'll reply re. the timings when I've worked out the details. In the
>meantime, here are the MAGICC global-mean temperatures simulated
>after tuning to the AR4 GCMs. For each SRES scenario, you'll find 3
>files (for climate-carbon cycle feedback strength low, default and
>high), and within each file you'll find multiple columns, one for
>each GCM that was tuned to.
>Let me know if anything isn't clear,
>Dr Timothy J Osborn, Academic Fellow
>Climatic Research Unit
>School of Environmental Sciences
>University of East Anglia
>Norwich NR4 7TJ, UK
>phone: +44 1603 592089
>fax: +44 1603 507784
>No virus found in this incoming message.
>Checked by AVG - http://www.avg.com
>Version: 8.0.138 / Virus Database: 270.4.10/1549 - Release Date:
>7/12/2008 4:31 PM