date: Thu Dec 1 10:42:18 2005
from: Keith Briffa <k.briffaatXYZxyz.ac.uk>
subject: Re: ENSO and Hadcm3
to: "Rob Wilson" <rob.wilsonatXYZxyzac.uk>, "Brohan, Philip" <philip.brohanatXYZxyzoffice.gov.uk>
Rob et al
this looks good on the surface and would have been well worth including - IF - as you say
, the Cook and the Mann reconstructions had been independent. I do not believe they are.
Surely Mann included the Texas/Mexico tree-ring data (produced largely by Stahle) in
predictors - perhaps in the form of major PC amplitude series ? We need to check this . I
am sure these are the major contributer to Ed's reconstruction .
At 14:26 30/11/2005, Rob Wilson wrote:
Hi Philip, Tim and Keith,
if you have a minute, any comments on my musings below would be greatly appreciated.
in trying to address some of Mike Evans' comments, I am going to add an extra section to
the paper - essentially comparing the reconstruction and models to ENSO.
Spectral analysis (MTM) of the coral recon, and the ALL run for HADCM3 identifies
significant (99%) spectral peaks at secular (>~90 yrs) scales and within the ENSO
bandwidth. ECHO-G does not show any ENSO equivalent spectral peaks at this high
confidence limit, although some spectral peaks are identified at the 95% level.
In general the MTM spectra for the coral recon and HADCM3 ALL are surprisingly similar -
However, after high pass filtering the time series with an 8 yr Gaussian filter, there
is little coherence between the coral recon and HADCM3 (r = -0.03)
I chose Ed's NINO3 TR based reconstruction for comparative analysis - he reconstructed
Dec-Feb NINO3 SSTS back to 1408. This reconstruction is completely independent to the
coral recon. I thought this better than Mann's NINO3 recon as it included some coral
Anyway, after high pass (8 yr) filtering, over the 1870-1978, the correlation between
instrumental annual tropical SSTs and Dec-Feb NINO3 SSTs = 0.76.
The correlation between the coral recon and TR NINO3 recon over the same period = 0.47.
Weaker, but the series are after all independent.
Again, there is no coherence between HADCM3 and Ed's NINO3 recon.
So - if I interpret these results correctly, HADCM3 does portray variability at the ENSO
time-scale, but this variability has NOT been 'tuned' to the real world - i.e. the model
all run does not correlate with reality. I am sure I am not the first to notice this -
is there any relevant references?
Professor Keith Briffa,
Climatic Research Unit
University of East Anglia
Norwich, NR4 7TJ, U.K.