cc: k.briffa@uea.ac.uk, p.jones@uea.ac.uk, paleo@cox-internet.com date: Tue, 10 Apr 2001 09:31:40 -0500 from: tom crowley subject: Re: Briffa et al reconstruction 2001 to: Tim Osborn Dear Tim, Keith, and Phil thanks very much for your comments - I can only answer briefly now - more later. I scaled the model, the revised Crowley-Lowery (CL6), and the decadally smoothed Briffa record to the mid-20th century temperature increase. The extent to which this is justifiable I am still mulling over. BUT if it has some legitimacy, the model explains 88% of the variance in the instrumental record and best fit sensitivity is somewhere between 3.0-3.5 C. Some of the surface proxies can be interpreted in terms of larger changes than Mann et al because they show early 20th century temperatures warmer than the early 19th and 17 centuries (alpine retreat data also show some warming over this interval). If CL and Brf are scaled in the same way as the model they provide some support for a greater surface sensitivity too. These are topics I plan to raise at the Virginia meeting - at this stage I am by no means wedded to the alternate interpretation but I do think it worthwhile to ponder the implications of an alternative fit (this was suggested to me by Gabi). Attached are two ps files - one showing the model fit to the instrumental record - in this case I took basically the same simulation that fit Mann so well pre-1850 and just rescaled it. I also show the borehole on it. In the second figure I show the comparison of the model with CL6 and scaled Briffa et al. There is much more variance in the Briffa record and the general trends are the same although the correlations are weaker. For the entire record the model can explain 70% of the variance in the CL6 record, spliced to Folland after 1950 (due to dropoff of proxy data). The model explains 42% of the decadally smoothed variance in the Briffa et al record over the interval 1407-1955. I look forward to any further comments any of you may have, as I value them very much. Regards, Tom >At 13:56 26/03/01 -0600, tom crowley wrote: >>Tim, thanks for the data. attached is a ps file comparing the Briffa et al >>2001 time series rescaled to NH mean annual temp using the 1881--1960 >>interval for fitting (r = 0.81). The UKMO data set refers to the recent >>paper submitted to GRL. > >I've never tried to reconstruct NH mean annual temp from it, as I've always >used only Apr-Sep and land only data (and usually extra-tropical only). >The r=0.81 seems very high. Was this for decadally smoothed data? If so >it might be somewhat overfitted because of the dominant trend, though we >have ourselves tried fitting separately at interannual and interdecadal >time scales and find some significant differences in regression slope. >Also, when you say "rescaled", do you mean regressed, or do you mean scaled >so its variance matches that observed? Again the latter might explain the >larger scaling that you have obtained. > >>Also shown is a stretched version of a modified Crowley-Lowery time series >>(termed CL6) - the stretch refers to the rescaling required to agree with a >>model best fit to the instrumental part of the record (which in turn >>implies a sensitivity of about 3.0 C for a doubling of CO2) > >So are you saying that the agreement between the Briffa et al. series >scaled to match observed temperatures with the CL series scaled to match >your model output implies that the scaled Briffa et al. series is likely to >be a good match to the model output? If so, then of course it would be >nice to actually see model output versus scaled Briffa et al. > >>Any comments? If there is some agreement as to the legitimacy of this >>exercise several shortcomings in the Crowley 2000 study (the instrumental >>record disagreement), then some important advances in understanding. That >>still leaves the Mann et al record outside the fold but I continue to worry >>whether his calibration, which extends to 1980, results in a "flattening" >>of the regression line and overestimate of paleotemperatures that you >>people discussed in one of your Nature papers. > >The legitimacy question revolves around exactly what you've done to the >Briffa et al. record to obtain this scaling - see my questions above. If >that rescaling is legitimate then the results, as you say above, have >important implications for (i) believing the model output, (ii) reconciling >with the borehole data, and (iii) comparison of the various proxy-based >reconstructions. Note that Mann et al. don't use any tree-ring density >series, which are the ones that are most affected by the recent "decline" >in high latitude density (relative to warm-season temperature) - therefore >extending his calibration through to 1980 probably wouldn't cause a >flattening of the regression line. > >Hope these comments are useful. > >Tim > > >Dr Timothy J Osborn | phone: +44 1603 592089 >Senior Research Associate | fax: +44 1603 507784 >Climatic Research Unit | e-mail: t.osborn@uea.ac.uk >School of Environmental Sciences | web-site: >University of East Anglia __________| http://www.cru.uea.ac.uk/~timo/ >Norwich NR4 7TJ | sunclock: >UK | http://www.cru.uea.ac.uk/~timo/sunclock.htm Attachment Converted: "C:\EUDORA\Attach\Forc.CL6.Brfscl.ps" Attachment Converted: "C:\EUDORA\Attach\Inst.fit.borehole.ps" Thomas J. Crowley Dept. of Oceanography Texas A&M University College Station, TX 77843-3146 979-845-0795 979-847-8879 (fax) 979-845-6331 (alternate fax)