cc: k.briffa@uea.ac.uk date: Fri, 12 Apr 2002 12:16:57 +0100 from: Tim Osborn subject: Re: Fwd: RE: Recent NH reconstruction to: Edward Cook At 12:55 08/04/02, Edward Cook wrote: >Some quick questions, comments, and a request. First, when you say that >you recalibrated all of the series shown in your windy perspectives bit, >what time period did you recalibrate all the series over? 1881-1960. >Was it the same for all series? Exceptions (as mentioned in Briffa et al., 2001, JGR paper on our age-banded reconstructions) are: (1) because Overpeck et al. provided only 5-yr mean values, the instrumental data were converted to matching 5-yr means. (2) because Crowley and Lowery's series had reduced high-frequency variability (due to their use of some lower-resolution proxies, and the way they combined these with the higher-resolution ones), the instrumental data were smoothed with a decadal filter. Plus we omitted the period 1901-1919 from the calibration. Crowley and Lowery omitted a similar (perhaps a bit longer?) period when they did their calibration. I shortened and moved the omitted period to 1901-1919 because that seemed better when calibrating against warm-season temperatures (which is what we put into the 2001 JGR paper). In our recent Science piece we calibrated against annual temperatures, but I still left it as omitting 1901-1919 from the Crowley calibration (because that's what my program does!). Omitting a period because the proxy and instrumental data disagree is not a good thing to do. I think Crowley and Lowery make some weak arguments to justify it, though I'm not convinced by them. But they did omit some data, so we do the same (with some concern). I'm also a little hesitant about being too critical, because one might view our use of a calibration period that ends in 1960 as being a similar thing - we omit the post-1960 period because of the apparent decline in high-latitude tree-ring density! Anyway, for both Overpeck et al. and Crowley and Lowery, the calibration is effectively fitting against the low-frequency trend. There's no need to express the implications of this, since you covered that already in your reply to Mann & Hughes. >Just to be clear here, I thoroughly dislike the pre-1900 instrumental >data. That pre-1900 warming back to 1860 or so is very suspicious. No >large-scale pure proxy estimates have ever been able to reproduce it. Only >the MBH series does because it includes instrumental data. I have similar concerns, but little expertise to be authoritative in attributing it to changes in thermometer exposure. It, as you probably know, is particularly noticeable in the warm season and over land. Given the location of the, e.g., the density tree-ring network I really feel it is preferable for us to reconstruct land, warm-season temperatures rather than going for the entire land & marine, annual NH mean. Yet it choosing a more optimal and defensible thing to reconstruct, we are penalised by the early summer warming. Given the apparent density decline, we're being squeezed from both ends of the record - and 1881-1960 seemed the best period to choose. >Thus, recalibrating the proxies using pre-1900 observed data may be >biasing the results. That is why I did not use the pre-1900 data in my >calibration. Phil would probably disagree strongly with me here, perhaps >you and Keith too, but that is my opinion. Also, how did you express all >series as anomalies from the observed 1961-1990 mean when not all series >extended out to 1990 (e.g. Briffa1, Mann)? I haven't checked, but I'd guess that *none* of the calibrated series have a zero mean over 1961-1990. But their mean over the calibration period (1881-1960, except for Crowley & Lowery - see above) will be forced by calibration to match the mean of the instrumental temperatures over the same period. Since the instrumental temperatures are expressed as deg C anomalies from the 1961-90 mean, the calibrated reconstructions have the same units. >I honestly dislike such short anomaly periods in any case, especially when >the data are trendy. It unnaturally distorts the visual expression of the >data (my opinion anyway). Yes - I've been playing around with calibrating the proxy records against each other (rather than against instrumental data) over earlier and much longer periods - it really highlights the similarity of the decadal-century variability (in many cases due to the common proxy records, of course). >Finally, can you send me the unfiltered observed land-only data that you used? File is attached. Best wishes Tim Attachment Converted: "c:\eudora\attach\ed_landtemp_20N90N.amjjas" 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