date: Tue, 21 Jul 1998 17:40:38 +0100 (BST)
from: Jonathan Gregory <jmgregory@meto.gov.uk>
subject: derivation of headline results
to: scenarios@meto.gov.uk

Dear Colleagues

I believe that it is important to clarify how the "headline" figures for
climate change appearing in the TAR and its summaries should be derived from
atmospheric concentrations.  The choice of approach will affect the contents of
more than one chapter.  I am not addressing the difficult issues of emissions
scenarios and how they are converted to concentrations, which are also being
debated in this email list. I am particularly writing to the proposed LAs of
the climate-change and sea-level chapters.

One possibility would be to restrict the TAR to assessment of results for
climate change and sea-level rise which come from coupled GCMs. This would be a
clean and attractive approach if the GCMs involved had all been run with the
same concentration scenarios. However, it would not give the complete range of
uncertainty because most GCMs are unlikely to have been run with the extreme
concentration cases. If it is the assessment of the TAR that the extant GCMs do
not cover the complete plausible ranges of climate sensitivity and ocean heat
uptake, those uncertainties would also not be fully represented.

Because of these limitations, it is probable that some means will be needed
for calculating results for outlying scenarios and sensitivities. In the SAR,
simple climate models were used for this. I think the role of simple models
in the TAR should be discussed now.

Simple models of aspects of the climate system are very useful for
investigating and understanding how things work. Such models should, I think,
be discussed in the TAR along with the processes they are used to investigate.
Simple models of the whole climate system (such as the two used in the SAR) are
also useful in their own right for investigating the effect of different
concentration scenarios, for instance. However, I think that a question which
has to be clearly addressed in the TAR is whether a simple model is in any way
scientifically superior to a coupled GCM for doing this job (setting aside the
practical issue of whether you can afford to use a GCM). The answer to this
question will determine how simple models are presented.

Question 1:

(a) If GCMs are the principal tools, producing results which are more reliable
in detail than those from any simple model, then simple models are needed only
as tools for EXTRAPOLATING or INTERPOLATING between GCM results, to estimate
the effect of different scenarios or sensitivities.

(b) If the assessment is that some simple models offer independent climate
predictions to be treated on the same footing as coupled GCMs, then simple
models should be discussed at much greater length in the model assessment and
model predictions chapters than they were in the SAR. An assessment of all
available simple models and their results should be given.

My view is (a), which I think is consistent with the IPCC simple models
paper. If (a) is the view taken by the authors of the relevant chapters, then I
argue that what we are seeking to do the job is a NUMERICAL TOOL. The test of
the tool is that it can reliably reproduce the results of several coupled GCMs,
both for temperature change and thermal expansion, for different kinds of
scenarios (e.g. 1%, historical forcing, stabilisation). This is a stringent
requirement. I do not think that a 2D climate model is necessarily preferable
to a 1D climate model for this task. In fact, the more elaborate the simple
model - the more physics it has got in it - the less amenable it will probably
be to being tuned to reproduce GCM results. It is essential that this
calibration can be demonstrated to be good, because in view (a) it is the GCMs
which are basically providing the results. The tool must not be allowed to go
its own way. It is not necessary to interpret the tuning parameters of a simple
model in terms of the physics of the real climate system when it is being used
for this purpose. It is not necessarily the case that a simple climate model is
the best choice anyway.

Question 2:

If this view is followed, the calibration of the tool will be needed for both
the climate-change and sea-level chapters. Where is the best place to put it?

(a) It could be in the climate-change chapter, which is earlier, in which case
we will have to take care that the climate-change and sea-level chapters
fit well together.

(b) It could be in the scenarios chapter, which may be dealing with this
kind of issue anyway.

(c) It could be in an annex or additional "results" chapter, in which case the
climate-change and sea-level chapters would present GCM results only. This
would have the drawback that the headline results of the TAR would look rather
like a footnote in the main text, even though they would feature prominently in
the summary. However, that's what was done in the 1992 report.

>From the point of view of clarity, I think (c) would be best.  It would have to
be presented as a main conclusions chapter of the TAR, resting on the GCM
results which had earlier been extensively discussed.

Question 3:

How many such tools are needed i.e. how many simple climate models or
equivalent? I think more than one should be investigated, to see how well they
fit the GCMs. If they all fit well, and their extrapolations to outlying
scenarios agree closely, only one need be presented in the TAR.  If there is a
range of results, this gives a measure of the additional uncertainty in the
procedure of extrapolation, and should be reported.

Thanks for reading all this!

Jonathan Gregory  jmgregory@meto.gov.uk  +44 1344 854542  fax +44 1344 854898
Hadley Centre, Met. Office, London Road, Bracknell, Berks., UK. RG12 2SY
