date: Fri, 28 Apr 2006 16:14:10 +0100 from: Tim Osborn subject: Standardization uncertainty for tree-ring series to: Keith Briffa ,Tom Melvin Tom, copied below is an email from Philip Brohan that describes his ideas about quantifying uncertainty. Keith and I probably need to talk with you about some specific issues (i.e., what type of RCS), but I thought I'd forward the basic plan anyway so that you can think about the logistics of doing it. If you calculated the RCS curve and then *subsequently* did a bootstrap calculation (where you sample with replacement, as described below) using the standardized indices to develop regional chronologies, then you would probably get a similar uncertainty to what you already obtained using the standard deviation of the individual core values. But the difference here is that the bootstrapping is done *prior* to the creation of the RCS curves, and thus the RCS curves will be different for each bootstrap iteration. The idea being that this will help to quantify uncertainty in the RCS curves themselves, and how this carries over into chronology uncertainty. What do you think, both about whether this really will quantify RCS uncertainty, and whether this is easy to implement? Cheers Tim At 16:02 26/04/2006, philip.brohan@metoffice.gov.uk wrote: >Keith, Tim. > > At our meeting last Wednesday I agreed to specify exactly what needed >to be done to make uncertainty estimates for standardisation of the >tree-ring data. > > Suppose we are making a proxy series from n cores. From those n cores >we can make an RCS age correction curve, and a mean proxy series (the >average of the cores after applying the age correction curve to each >one?). These are the best-estimate values for the age-correction curve >and the proxy series. > > We also need bootstrap estimates of the age correction curve and the >mean proxy series. To make a bootstrap estimate: sample, with >replacement, from the n cores until you have a set of n samples. (Some >of the cores will be in this sample once, some several times, and some >not at all). From this set of n samples, make an age correction curve >and a mean proxy series as before. These are the bootstrap estimates. > > We need a lot of bootstrap estimates. I'd like 1000 - 100 will probably >do at a pinch. So please can you make these and send me the 1001 age >correction curves and 1001 mean proxy series. > > I will do something similar with the instrumental series, and we can >then make bootstrap estimates of the regression uncertainty and the >uncertainty in the reconstructed temperatures. > >Cheers, > > Philip > >-- >Philip Brohan, Climate Scientist >Met Office Hadley Centre for Climate Prediction and Research >Tel: +44 (0)1392 884574 Fax: +44 (0)1392 885681 >Global climate data sets are available from http://www.hadobs.org Dr Timothy J Osborn Climatic Research Unit School of Environmental Sciences, University of East Anglia Norwich NR4 7TJ, UK e-mail: t.osborn@uea.ac.uk phone: +44 1603 592089 fax: +44 1603 507784 web: http://www.cru.uea.ac.uk/~timo/ sunclock: http://www.cru.uea.ac.uk/~timo/sunclock.htm