cc: ddokken@usgcrp.gov
date: Thu, 30 Jan 1997 12:18:00 -0500
from: ddokken@usgcrp.gov (Dave Dokken)
subject: Re: Regional Climate Projection Piece
to: Mike Hulme <m.hulme@uea.ac.uk>

Thank you Mike. Since you never saw the figs/tables, I faxed them to your
attention.

dave

>Dave,
>
>Here are some comments on the IPCC Special Paper on Regional Climate
>Impacts.  I have copied these to Richard also and have put a manuscript in
>the post which you might care to pass on to the authors of the IPCC
>manuscript for me.
>
>Thanks,
>
>Mike
>
>
>Simulation of Regional Climate Change with Coupled Climate Models and
>Regional Modelling Techniques
>
>Some comments by Mike Hulme
>Climatic Research Unit
>UEA,  Norwich  NR4  7TJ
>(email:  m.hulme@uea.ac.uk)
>
>
>I found the paper clear and well-written and it should form the basis of a
>useful introduction for the Special Report.  I have not seen the Figures so
>I cannot comment on these in detail.  My additional comments are below:
>
>p.1, Regional Simulations .... 1st para:  Figure 1: it is not strictly
>legitimate to compare change fields from eq. and tr. experiments at the time
>of CO2 doubling (because of the difference between equilibrium and realised
>change) or between warm-start and cold-start tr. experiments at the time of
>CO2 doubling (because realised warming is a function of the historic forcing
>which may or may not have been included).  I have not seen a list of the
>experiments in Figure 1 so I do not know how far this is a problem.  But be
>careful.  Even if it is not a problem, the point should be made that DC at
>time of CO2 doubling in a tr. experiment is not equivalent to DC at time of
>CO2 doubling in an eq. experiment.  People used to dealing with eq. 2xCO2
>experiments may get confused here.
>
>p.2: the point should be made when comparing regional change patterns from
>different GCMs that some (maybe a lot) of the differences between
>experiments is due to interdecadal variability in the experiments
>(especially true of non-temperature variables).  A range of precip. changes
>for a region from -20% to +20% is not a robust measure of the true
>disagreement amongst the models about the GHG rainfall signal since no
>allowance is made for interdecadal variability (i.e., signal to noise
>ratios).  Several simulations from the same model and same forcing would
>also yield a range of regional precip. changes which might be just as
>large!!  Little work has been published on the analysis of GHG signals from
>ensemble experiments, but see Cubasch et al. (1995) for a helpful example
>(also a manuscript in preparation by John Mitchell).  I think this point is
>an important one to make (see also later).
>
>p.3:  final bullet:  this difference between the statistical significance of
>precip. and temp. biases is most important (see my point above).  A regional
>rainfall 'signal' of +10% from a single GCM simulation may be meaningless
>against the background noise.
>
>p.3:  I do not like your use of the term 'model sensitivity' (and elsewhere
>'climate sensitivity', e.g. line 17) to describe the regional DC from a
>model simulation.  This is dangerously confusing with the proper use of the
>term 'climate sensitivity' which is the equilibrium warming from an
>instantaneous doubling of CO2-equivalent.  'Biases' is fine for
>model-observed differences; perturbed minus control differences should be
>called 'regional climate changes', not sensitivities.
>
>p.3, lines 12/13:  the sentence, ' ... the range of simulated scenarios of
>the model regional biases were still large ....' is muddled.  I cannot
>discern your meaning.
>
>p.4, section on aerosol and GHG signals:  I think again the point about
>signal to noise ratios should be made.  Just because an aerosol simulation
>changes the sign of precip. change for a region, does not mean that this is
>at all significant.  If the rainfall change for a decade (or 30-year period)
>switches from +5% to -5% when aerosols are included means nothing if the
>natural 10-30 year rainfall variability is, say, 20% (see my manuscript
>Hulme & Brown).
>
>p.5, 'Precipitation' section:  here it is stated (twice) about a general
>reduction in precip. over southern Europe, but on p.2 all models agreed
>about a winter increase in precip. over the same region!  Which is correct?
>
>p.5: I think this point about switching of rainfall change over Asian
>monsoon in the aerosol experiments needs handling with some care.  While it
>may be true for the particular domain chosen here (5-30N, 70-105E), if one
>looks within this domain or else looks at surrounding regions, one does not
>necessarily get the same result.  Here is a case where the headline rainfall
>change number is highly sensitive on the region defined.  For example, in
>HADCM2 experiment over the Indian land area only, both GHG and aerosol
>experiments yield a decrease in rainfall (see Hulme & Brown).
>
>pp.5/6, soil moisture section:  I think the point should be made that many
>environmental modellers prefer to calculate soil moisture themselves using
>GCM output as an input into their (often) more sophisticated models of soil
>and the water cycle.  GCM soil moisture schemes are very crude!  Relying on
>GCMs for soil moisture changes is not often a good idea.
>
>p.7, statistical downscaling:  a word of caution is needed here about
>downscaling.  Just because these methods may yield more spatial detail
>and/or actual point data does not necessitate that these scenarios are in
>any way 'better' than the GCM-derived ones.  Downscaling methods still rely
>on GCM output to drive them and, furthermore, most downscaling methods make
>the big assumption that synoptic-weather relationships will remain unchanged
>between the past and the future.  The impression must not be given that
>downscaling will solve all scenario difficulties (see below).
>
>p.7, line -11:  change 'Limited applications ...' to 'A few applications ....'
>
>pp.7/8, regional modelling:  very few climate change experiments with
>regional models using more than 5-10 years of simulation.  I do not know of
>a high resolution climate change experiment for Africa (p.8, line 12).
>There have been high res. experiments that include aerosols (p.8, lines
>14/15) - the Hadley Centre have done this.
>
>pp. 7/8: time slice experiments should be mentioned here (see Cubasch et
>al., 1995).
>
>p.9, line -5:  'inherent predictability'  this statement may well be true,
>but it depends what you mean by it.  See my earlier comments about
>signal/noise ratios.
>
>p.10, line 3:  a good example where 'model sensitivities' should be replaced
>by 'regional climate changes'.
>
>p.10, lines 7/8:  dangerous to imply that high resolution models yield high
>accuracy scenario!!!  They may well yield more faithful control simulations
>of regional climate, but their 'accuracy' re. future scenarios still depends
>on assumptions about what drives the climate (i.e., what emissions scenario)
>and about how good the driving GCM is!  If your comment about inherent limts
>to climate predictability is true then it doesn't matter what resolution the
>climate model is!!! Even 1km resolution can give you appearance of excellent
>control climate but says nothing about the accuracy of any scenario.
>
>These comments above also apply to the last paragraph.  It seems you to me
>that you are too upbeat about what high resolution models can do for us.
>Climate prediction will still be elusive because:
>
>a)  we do not know future forcing
>b)  we do not know the climate sensitivity (1.5 to 4.5C)
>c)  there are inherent limitations on climate predictability which are
>resolution independent.
>
>
>
>Some Other References:
>
>
>Airey,M., Hulme,M. and Johns,T. (1996)   Performance of the Hadley Centre
>GCM in simulating precipitation over land areas   Geophys. Res. Letts.,  23,
>1657-1660.
>Cubasch,U., Waszkewitz,J., Hegerl,G. and Perlwitz,J. (1995)   Regional
>climate changes as simulated in time-slice experiments   Climatic Change,
>31,  273-304
>Hewitson,B.C. and Crane,R.G. (1996)   Climate downscaling: techniques and
>application   Climate Research,  7,  85-95.
>Hulme,M. and Brown,O. (1997)   How likely is tolerable regional climate
>change?   Unpublished manuscript,  UEA,  Norwich.  (copy mailed to Dave Dokken)
>
>Risbey,J.S. and Stone,P.H. (1996)   A case study of the adequacy of GCM
>simulations for input to regional climate change assessments   J. Climate,
>9,  1441-1467.
>Whetton,P., England,M., O'Farrell,S., Watterson,I. and Pittock,B. (1996)
>Global comparison of the regional rainfall results of enhanced greenhouse
>coupled and mixed layer ocean experiments: implications for climate change
>scenario development   Climatic Change  33,  497-519.
>
>
>
>
>
>At 12:33 15/01/97 -0500, you wrote:
>>Dear Dr. Hulme -
>>
>>Richard Moss asked that I forward this draft chapter for the IPCC Special
>>Report on the Regional Impacts of Climate Change to your attention. Since
>>it is now a Special Report, we can bring in post-SAR material, and you may
>>have an opinion on how to augment the piece.
>>
>>Any observations should be forwarded to both Richard and my attention,
>>since he's on travel for the next 4 weeks solid! For easy reference, his
>>e-mail address is rmoss@usgcrp.gov.
>>
>>I'm merely going to attach the MSWord 6.0 file as a quick fix. If you
>>cannot read it, get back to me and I can pipe it from our server as an
>>ASCII, convert it to your preferred word-processing package, or rely on the
>>tried-and-true fax.
>>
>>Thanks for your help,
>>
>>
>>
>>Attachment Converted: c:\eudora\attach\Regional.doc
>>
>>Dave Dokken
>>Project Administrator
>>IPCC Working Group II TSU
>>Code YS-1
>>300 E Street, SW
>>Washington, DC  20546
>>+1.202.358.0507 (voice)
>>+1.202.358.4104 (fax)
>>
>>
>>
>-----------------------------------------------------------------------------
>Dr Mike Hulme                           tel: +44 1603 593162
>Climatic Research Unit                  fax: +44 1603 507784
>School of Environmental Sciences        email:  m.hulme@uea.ac.uk
>University of East Anglia               web site:
>http://www.cru.uea.ac.uk/~mikeh/
>Norwich  NR4  7TJ
>
>
>Mean temperature in C.England during 1996 was 0.3degC below the 1961-90
>average.
>The maximum temperature in Norwich:    Tuesday 28 January:  8.1degC.


