date: Tue Jul 18 12:43:48 2006
from: Tim Osborn <t.osborn@uea.ac.uk>
subject: Re: new fig 6.14
to: Fortunat Joos <joos@climate.unibe.ch>, Jonathan Overpeck <jto@u.arizona.edu>, Keith Briffa <k.briffa@uea.ac.uk>, Eystein Jansen <eystein.jansen@geo.uib.no>

   Hi all,
   thanks for the responses, Peck and Fortunat.
   I drafted the new figure 6.14 following as closely as possible the approach used for the
   original forcing/simulation figure (now 6.13).  This is why I smoothed all series and used
   a common anomalisation period for all curves across all panels.  It can greatly help to
   interpret why the simulated temperature responds in the way it does, because the zero (or
   "normal" level) is comparable across plots and because the strengths of different forcings
   can be compared *on the same timescale* as the simulated temperatures are shown.  And, for
   6.13, with so many different forcings and models shown, it would have been impossible to
   use unsmoothed series without making the individual curves indistinguishable (or indeed
   fitting them into such a compact figure).
   Now that the EMIC panels are separate from the original 6.13, we do have the opportunity to
   make different presentational choices.  But I think, nevertheless, that some of the reasons
   for (i) proportional scaling, (ii) common anomalisation period; and (iii) smoothing to
   achieve presentation on comparable time scales, that held for 6.13 probably also hold in
   6.14.
   However, I also appreciate the points raised by Fortunat, specifically that (i) it is nice
   to be able to compare the magnitude of the 11-yr solar cycles with the magnitude of the
   low-frequency solar variations; and (ii) that using a modern reference period removes the
   interpretation that we don't even know the forcing today.
   So we have various advantages and disadvantages of different presentational choices, and no
   set of choices will satisfy all these competing demands.
   One thing that I am particularly perturbed about is Fortunat's implication that to show
   smoothed forcings would be scientifically dishonest.  I disagree (and I was also upset by
   your choice of wording).  If it were dishonest to show smoothed data, then presumably the
   same holds for 6.13 (but its impossible to distinguish all the different volcanic forcings
   if shown unsmoothed), but also to every other graphic... should I be showing the EMIC
   simulated temperatures without smoothing too, so you can see the individual yearly
   responses to the volcanic spikes?  But annual means are formed from the temperatures
   simulated on the model timesteps, so we still wouldn't be showing results that had not been
   post-processed.  Most climate models, even GCMs, respond in a quasi-linear way, such that
   the smoothed response to unsmooth forcing is very similar to the response to smooth
   forcing.  So if we are interested in the temperature response on time scales of 30 years
   and longer, it seems entirely appropriate (and better for interpretation/comparison of
   forcings) to show the forcings on this time scale too, because the forcing variations on
   those time scales are the ones that are driving the temperature response (even though the
   forcing may be intermittent like volcanoes or have 11-yr cycles like solar).
   The choice of smoothing / no smoothing is not, therefore, anything to do with
   honesty/dishonesty, but is purely a presentational choice that can made accordingly to what
   the purpose of the figure is.  Here our purpose seems to be long-term climate changes,
   rather than response to individual volcanoes or to the 11-yr solar cycle.
   So the position is:
   (1) smoothing or no smoothing: there are arguments for both choices, though clearly I
   prefer smoothing and Fortunat prefers no smoothing.  I could make a figure which kept the
   smooth lines but put the raw annual histogram volcanic spikes underneath in pale grey, as
   Peck requested anyway (and possibly put the 11-yr solar cycles in pale brown underneath the
   smoothed brown solar series).  This would be a compromise but the main problem is that the
   scale of the largest volcanic spikes would far exceed the scale I am using to show the
   smoothed series (so the panel is not large enough to do this)!
   (2) pre-industrial or present-day anomalisation reference period: again there are arguments
   for both choices.  Whatever we choose, I firmly believe it should be the same for *all*
   curves in this figure (which can make a dramatic difference).
   (3) exaggeration of solar scale or proportional vertical scales: this is the one that I
   have the firmest opinion about.  I see no reason to exaggerate the scale of the solar
   forcings relative to volcanic or anthropogenic forcings.  The difference between the
   forcings looks clear enough in the version of the figure that I made.  Exaggerating it will
   wrongly make the Bard 2.5% case look (at first glance) bigger than the anthropogenic
   forcing, and make it look more important than volcanic forcing.
   I'll hold off from making any more versions till decisions are made on these issues.
   Cheers
   Tim
   At 09:01 18/07/2006, Fortunat Joos wrote:

     Hi Tim and co,
     Thanks for the figure. I like the figure showing the model results and the general
     outline/graphic style.
     However, I am concerned about what is shown in the forcing figure.
     1) Volcanic panel: I strongly believe that we should show what was used by the model and
     not some 40 year smoothed curves for volcanic forcing or any other forcing. So please
     use the original data file. Scientific honesty demands to show what was used and not
     something post-processed.
     2) solar panel:
     2a) We must show the Wang-Lean-Shirley data on the original resolution as used to drive
     the models. In this way, we also illustrate the magnitude of the 11-yr annual cycle in
     comparison with the background trend. The record being flat, apart from the 11-yr cycle,
     during the last decades is a reality.
     2b) Do not apply any smooting to the Bard data. Just use them as they are and how they
     were published by Bard and used in the model.
     2c) It is fine to supress the Bard 0.08 case after 1610 (not done in my figure version)
     2d) the emphasis of the figure is on the solar forcing differences. So, please show
     solar somewhat overproportional in comparison to volcanic and other forcings.
     3) other forcings: again no smoothing needed here. It would be hard to defend a double
     smoothing.
     4)- normalisation of solar forcing to some period mean. If the different solar forcings
     disagree for today as in your option, we may send the signal that we do not even know
     solar forcing today.
     Thus, I slightly prefer to have the same mean forcing values for all solar records
     during the last few decades as shown in the attached version. However, I also can see
     some arguments for other normalisations.
     To illustrate points 1 to 4, I have prepared and attached a version of the forcing
     panel.
     other points
     - Your choice of colors is fine
     - time range 1000-2000 AD is fine
     - suggest to remove the text from the y-labels except the units W/m2.
     Sorry for this additional comments coming a bit late. However, I did not realise that
     you planned to smoothed the model input data in any way.
     With best wishes,
     Fortunat
     Tim Osborn wrote:

     Hi Peck, Eystein and Fortunat,
     I've drafted two versions of the new fig 6.14, comprising a new panel showing the
     forcing used in the EMIC runs, plus the old fig 6.13e panel showing the EMIC simulated
     NH temperatures.  Keith has seen them already.
     First you should know what I did, so that you (especially Fortunat) can check that what
     I did was appropriate:
     (1) For the volcanic forcing, I simply took the volcanic RF forcing from Fortunat's file
     and applied the 30-year smoothing before plotting it.
     (2) For the solar forcing there are 2 curves.  For the first, I took the Bard 0.25%
     column from Fortunat's RF file.  For the second, I took the Bard 0.08% column from
     Fortunat's RF file from 1001 to 1609, and then appended the WLS RF forcing from 1610 to
     1998.  Then I smoothed the combined record.  NOTE that for the Bard0.25%, the line is
     flat from 1961 onwards which probably isn't realistic, even though that is what was used
     in the model runs.
     (3) For the "all other forcings" there are 2 curves.  For the first, I took the CO2
     concentrations provided by Fortunat, then used the "standard" IPCC formula from the TAR
     (in fact the first of the three options for CO2 in IPCC TAR Table 6.2) to convert this
     to a radiative forcing.  I then added this to the non-CO2 radiative forcings data from
     Fortunat's file, to get the total radiative forcing.  For the second, I replaced all
     values after 1765 with the 1765 value (for the natural forcings case).  Then I smoothed
     the combined record (as in fig 6.13c, I only applied a 10-year smoothing when plotting
     the "all other forcings", because it is fairly smooth anyway and using a high smoothing
     results in lower final values when there is a strong trend at the end of a time series).
     Now, some comments on the figures themselves (please print them and refer to them when
     reading this):
     (1) File 'chap6_f6.14_option1.pdf' is strongly preferred by Keith and me.  This shows
     the three forcing components separately, which helps with understanding the individual
     causes of specific warming and cooling periods.  I have managed to reduce the size of
     this considerably, compared to the equivalent panel in fig 6.13, because with only a few
     series on it I could squeeze them together more and also reduce the range of the
     vertical axes.
     (2) Although we don't prefer it, I have also made 'chap6_f6.14_option2.pdf' which is
     even smaller by only showing the sum of all the forcings in the top panel.
     Which version do you prefer?  Please let me know so I can make final changes only to the
     preferred version.
     Some more comments:
     (1) Fig 6.14b was originally Fig 6.13e.  When it was part of that figure, the colour bar
     showing the shades of grey used to depict the overlapping ranges of the published
     temperature reconstructions was only on Fig 6.13d.  Do you think I should now also add
     it to the EMIC panel (6.14b), now that it is in a separate figure?  It will be a bit of
     a squeeze because of the legend that is already in 6.14b.
     (2) Another carry over from when 6.14b was part of 6.13, is that the time range of all
     panels had to match (900-2010).  Now that the EMICs are in a separate figure, I could
     start them in year 1000, which is when the forcing and simulations begin.  Unless you
     want 6.13 and 6.14 to remain comparable?  Again please comment/decide.
     (3) I wasn't sure what colours to use for the forcing series.  In option 1, the volcanic
     and other forcings apply to all runs, so I chose black (with thick/thin used to
     distinguish the "all" forcings from the "natural-only" forcings (basically the thin flat
     line in "all other forcings).  The cyan-green-blue runs used strong solar forcing, so I
     used blue for that forcing.  The red-orange-brown runs used weak solar forcing, so I
     used brown for that forcing.  Sound ok?
     Sorry for the long email, but I wanted to get everything explained to avoid too many
     iterations.
     Please let me know your decisions/comments on these questions, or on any other aspects
     of the new figure.
     Cheers
     Tim
