date: Wed Jun 16 16:49:37 2004 from: Tim Osborn subject: Re: revised manuscript to: "Michael E. Mann" , Scott Rutherford , Bradley Raymond , Hughes Malcolm , Briffa Keith , Jones Phil , mann@virginia.edu Hi Mike, a few follow-ups to some of your points: At 14:50 16/06/2004, Michael E. Mann wrote: 1. In discussing how spatial reconstructions are averaged to get hemispheric means, we just meant that, not that these spatial reconstructions were the results of CFR approaches. I believe it is true that Briffa et al, 1998a does average gridpoint reconstructions to get a hemispheric mean, right? No, you're wrong: it is NOT true that the "NH" series in Briffa98a was the average of gridpoint reconstructions. We averaged uncalibrated MXD series into regions, and then the regions into a single series. That was then calibrated by linear regression against the temperature target. So Briffa et al. 1998a falls into the second group - i.e. those that go direct to the target index (rather than to a field first, from which the target index is subsequently computed/averaged). The slightly confusing aspect is that we also showed maps in Briffa et al. 1998a, but these were uncalibrated (just tree-density anomalies). Our first attempt at actually doing a spatial field reconstruction with the data is (finally) written up in Osborn et al. - though we've shown maps that result from this work in various other papers. It's not the ideal order to write these things up, but that's just how it's happened due to various pressures to do various pieces of work. I just hope that the Osborn et al. paper is accepted without requiring changes to the reconstruction method - if that were to happen then other papers (including Rutherford et al.) that have already used the results would be inconsistent. As I said, it's not an ideal order to do things, but if Osborn et al. is accepted then it'll be ok! 2. Re, McINtyre/McKitrick: they didn't identify any "errors" at all. Good! 4. I believe that including weights in the hybrid approach that are proportional to the fraction of variance resolved by the indicator is essential. In fact, we know that we get completely wild results if we don't do that... I understand that now. Thanks for explaining. I think a sentence needs to be added to clarify this point... Yes - unless its in there already and I missed it!. We should say (1) that different proxies are given different weights in the low and high-frequency reconstructions, and (2) how those weights were computed. From what Mike says I infer that the answer to (2) is: the weight used was based on the ratio of the series' variance that was in the <20-year band to the variance in the >20-year band. So, if a series has a strongly red spectrum, it's relative weight in the >20-year reconstructions will be much higher than it's relative weight in the <20-year reconstructions. And vice versa for a series with a blue spectrum. 5., However, I stand by the claim that they appear to use a solar forcing that is about twice that of Crowley's! In the talk they gave in Nice (Phil was there), they showed their solar constant varying between 1362 and 1368 W/m^2 over the past millennium! You're correct that they used a stronger forcing. I have their data - the range is 1362.22 in 1458 to 1368.55 in 1989 - making a range of 6.33 W/m^2 (equivalent to a forcing of 1.1 W/m^2). In the period with sunspot observations, their minimum is 1362.99 (late 1600s) making a range of 5.56 W/m^2 (equivalent to a forcing of 0.96 W/m^2). IPCC give the total solar irradiance change over this period a range of 3 to 5.5 W/m^2 (though acknowledge even bigger uncertainty both lower and higher). So the GKSS group have used a forcing larger than Crowley's (though perhaps not double - I'd have to check) and it is also at the top of the IPCC range (though still not outside the range of possibilities). Hope you find these figures of interest, even if not directly relevant to the current manuscript. Cheers Tim