From: "Rob Wilson" To: "Tim Osborn" Subject: Re: Emailing: Wilson et al. technical comment Date: Mon, 27 Feb 2006 14:28:29 -0000 Reply-to: "Rob Wilson" Cc: "rosanne" , Hi Tim, yes, we processed our own RCS chronology using Jan's Jaemtland data. I also agree that using Jaemtland or not would make little difference to the results. Rosanne is presenting at this NAS meeting on Thursday which McIntyre is obviously going to use as a forum to muddy the waters even further. He has given us a hard time about the use of Gaspe and the Polar Urals chronologies and their influence on the 'hockey stick' trend over the past 2 centuries. However, removing these series makes little difference to our results in the past few centuries. am just going through your e-mails w.r.t. the coral paper - it is a huge help thanks Rob ----- Original Message ----- From: [1]Tim Osborn To: [2]Rob Wilson Cc: [3]rosanne ; [4]K.briffa@uea.ac.uk Sent: Monday, February 27, 2006 2:23 PM Subject: Re: Emailing: Wilson et al. technical comment Thanks for the very clear answers Rob. We didn't use Jaemtland and you did, that is why McIntyre suggested that we disagreed. But in fact our reason for excluding it was not that it didn't correlate with temperature positively, but that we didn't even calculate a correlation because the RCS chronology series we received stopped in 1827 rather than 1978. It is true that the full set of core data from Jan Esper span the range 1107-1978, but the RCS chronology we received spanned the range 1316-1827 only - and this matches the replication diagram in Esper et al. ([5]http://www.sciencemag.org/content/vol295/issue5563/images/data/2250/DC1/1066208S2_me d.gif) which stops then for Jaemtland. Presumably you obtained the set of core data and did your own RCS processing etc., rather than using the Esper et al. RCS chronologies? Anyway, I think that clears up our supposed "differences" over Jaemtland, though do let me know if you have any more points to add. Our results would have been very little affected by including Jaemtland anyway! Cheers Tim At 09:58 25/02/2006, Rob Wilson wrote: >Moring Tim, >answers in red. > > >on a related matter, Science have forwarded me some >questions/requests from McIntyre about our paper that they'd like our >response to. One of them states that "D'Arrigo et al. (2006) have >reported directly opposite findings in respect to the correlation >between their RCS chronology and gridcell temperature for Jaemtland >and the two foxtail series." >I am not sure where he got that from. >We used Jaemtland - it is a good site. >We did not use the foxtail data for similar reasons for us not using >the Bristlecone pine data (see below) . > >We didn't give a correlation for Jaemtland so it is hard for you to >have obtained the "opposite of nothing"! But anyway, I wanted to ask >whether in fact your Jaemtland differed from the one we used. The >one we used should be the same as Esper et al., with data provided by >Ed Cook. You seem to be citing Naurzbaev and Vaganov (1999) for your >Jaemtland record which seems odd. And its start and finish years >differ from the series I got, so I'm guessing that the data are >different and thus there's no reason why different data would have >consistent correlations. Also, do you know what correlation and for >what season (annual-mean?) you got for Jaemtland? >We also used the Esper data. >The N+V reference is completely wrong. I checked with Rosanne. Not >sure how that got in. The N+V reference is actually for Taymir. >Apologies for that - hopefully there are no more mistakes like that. >Anyway, to clarify what we did to the data, here is an exert from >the report I wrote for Rosanne 2 years ago. > >"The data from this site were those utilised by Jan Esper for his >Science paper. After removing a few low correlated series, the final >data-set consists of 156 radii over the period 1106-1978. >Unfortunately however, the period 1292-1315 is represented by only >one radius and replication is only reasonable from the mid 14th century. " > >In the end, I used the period represented by 10 or more series - 1340-1978. >This should agree with the data you have. > >As for correlations with temperature, Jaemtland is OK. >Against the relevant local 5x5 Land CRU (version 1) grid, the STD >and RCS chrons correlate with the Jun-Sep season at 0.48 over the >1956-1970 period. No residual problems were found with this >relationship. All screening was done up to 1970 so that potential >divergence would not effect the screening process. In this situation >though, there was no divergence for the 1971-1978 period. > > >On your (D'Arrigo et al.) exclusion of the Boreal/Upperwright series, >it wasn't clear which (one or more) of the 3 reasons listed applied >to these: (1) no significant temperature correlation, (2) significant >precip correlation, (3) too far south. >I know that the temperature signal is debatable in such records, but >I seem to recall you saying that on the longer time scales they (and >I think you were referring to Boreal/Upperwright, but I may have been >mistaken) showed some agreement with the N. American series from this >recent paper, giving some support at least for a temperature >signal. Is my recollection correct? >As I said earlier, I did not look at the Foxtail data. >However, I have played with the BP data. >The sites I utilised are described in this extract. > >"Of the 10 Bristlecone pine chronologies sent to me, 3 chronologies >were identified to express a significant summer temperature signal >using correlation analysis against local gridded data. These three >sites also load upon the same principal component in a PCA using all >10 chronologies. These three sites are: Hermit Hill (N = 38; >1048-1983) and Windy Ridge (N = 29; 1050-1985) from Colorado and >Sheep Mountain (N = 71; 0 - 1990) from California (Figure 1)." > >The correlation of the STD and RCS chronologies against local >gridded July-Sep mean temperatures is 0.38 and 0.34 respectively. > >I have also showed you a comparative plot of the RCS chronology with >my North American average series and the comparison is pretty good >for most of the record and certainly there does not seem to be any >obvious inflation of index values in the 20th century. > >So - why did we not use this site: >well >(1) Steve Macintyre was kicking up a fuss about these data and we >felt that perhaps it might be opening us to criticism if we used them >(2) These data are have been reported to also show a precipitation >signal. I did some analysis on a site basis, but cannot find the >results. However, the precipiation signal in the 3 chrons used was >also weak. The temperature signal is stronger. This agrees with the >BP vs NA chronology comparison. >(3) As this was a low latitude site, then we would also need to >include other low latitude sites - e.g. from the Himalayas. Jan >would not let me use his data for this region, so in the end, we >decided to keep the data-set as high latitude as possible. Quebec, >Alps and Mongolia being the most southerly sites. > >I hope this answers your queries. Rosanne is presenting at the NAS >meeting next week, and we have been trying to address many of the >criticisms of Macintyre that he is posting on his blog. I think Jan >making his data available was probably bad timing. > >Rob Dr Timothy J Osborn Climatic Research Unit School of Environmental Sciences, University of East Anglia Norwich NR4 7TJ, UK e-mail: [6]t.osborn@uea.ac.uk phone: +44 1603 592089 fax: +44 1603 507784 web: [7]http://www.cru.uea.ac.uk/~timo/ sunclock: [8]http://www.cru.uea.ac.uk/~timo/sunclock.htm References 1. mailto:t.osborn@uea.ac.uk 2. mailto:rob.wilson@ed.ac.uk 3. mailto:rdd@ldeo.columbia.edu 4. mailto:K.briffa@uea.ac.uk 5. http://www.sciencemag.org/content/vol295/issue5563/images/data/2250/DC1/1066208S2_med.gif 6. mailto:t.osborn@uea.ac.uk 7. http://www.cru.uea.ac.uk/~timo/ 8. http://www.cru.uea.ac.uk/~timo/sunclock.htm