date: Mon, 20 Jun 2005 10:58:19 +0100
from: Tim Osborn <t.osbornatXYZxyz.ac.uk>
subject: RE: Palaeo reconstructions
to: "Jenkins, Geoff" <geoff.jenkinsatXYZxyzoffice.gov.uk>
well of course Defra can do as they like, but the 3 alternative figures all
(1) they all show results from Moberg et al. (2005) which we have
criticised in a response to Nature (which Nature have declined to publish,
on the grounds that it was too technical) and which Philip Brohan at the
Hadley Centre has also criticised. I have cc'd this to Philip in case he
wishes to share his criticisms with you. The basic complaint is that the
low resolution proxies, which entirely determine the long-term variations
shown by the Moberg series, have not been properly calibrated and thus the
long-term curve cannot be accepted as representing "degrees C".
(2) As the blue uncertainty ranges clearly show, this also has narrower
uncertainties near 1000 AD than in the "Little Ice Age". So if they are
unhappy that the figure you drew from my advice shows this same behaviour,
why would they be happy with the Moberg figures.
(3) The uncertainty ranges given for the MBH reconstruction look wrong to
me - compare with Fig 2.21 of IPCC TAR. Maybe it can be explained by
different smoothing/filtering or different reference period, or maybe it can't.
Finally I would add that, on reflection, there are actually four reasons
why the envelope narrowed near 1000 AD:
(1) The reason I gave before (fewer reconstructions from which to define
(2) There is some proxy data that is common to almost all the
reconstructions (thus they aren't all independent) and the proportion of
this common data is greater in the earlier parts of the series (thus there
is even less independence near 1000 AD).
(3) The envelope of reconstructions, nor any of the uncertainty ranges
associated with individual reconstructions, represent the full uncertainty
- in particular, very few include uncertainty due to the deterioration of
proxy data back in time (e.g., some tree-ring chronologies comprise of
fewer actual tree cores further back).
(4) Some aspects of the uncertainty are actually reduced in the 1000-1100
period compared with the "Little Ice Age". It seems that many of the
uncertainties and potential biases are compounded when reconstructing
values that are further from the calibration period mean (usually 20th
century mean). Thus the "Little Ice Age" values are perhaps truly more
uncertain than the 1000-1100 values, because the latter are nearer to the
20th century level. Therefore perhaps the narrowing of the uncertainty
range early on in these figures is correct (unless this point (4) is
dominated by points (1-3)).
Hope this helps. Perhaps Defra would like to fund us to give them some
advice on these issues (or even a small, critical study), instead of
getting the advice for free? :-)
At 14:50 16/06/2005, Jenkins, Geoff wrote:
>Sorry to did all this up again, but Defra arent happy with the figure we
>came up with on your advice, showing the envalope of all
>reconstructions. The reason is the point (3) you made below, ie it looks
>like uncertainties are smaller earlier on. Despite making this point in
>the text, Defra would prefer to use suomething else and suggest one of
>the figs from Moburg et al. Hence I have included a couple of
>alternatives with text attached (plus the old one for comparison). I
>would be really grateful if you could let me have your advice on which
>of these you think is the more acceptable.
>(3) It might also be worth saying that the apparent narrowing of the
>in the first 2 centuries isn't because of smaller uncertainties, but
>because of few reconstructions from which to define the "envelope".
Dr Timothy J Osborn
Climatic Research Unit
School of Environmental Sciences, University of East Anglia
Norwich NR4 7TJ, UK