BUCKLEY: We hope to extend our 14C chronology further, which may give us an answer to that question. Also, I expect to find similar temperature resolution at other sites in Tasmania next year.

GHIL: I can show the same peaks from an instrumental record [see Figure 1 in the essay introducing Atmospheric Modeling in Chapter 2]. On a power spectrum of a 300-year record of central England temperatures obtained by the singular spectrum analysis method you can clearly see two interdecadal and two interannual peaks. One is 25, near enough to your 30, and one 14.2, like your 15. The interannual peaks are 7.6, near your 8, and 5.2, matching your 5.2.

MYSAK: The 200-year time scale you passed over rather lightly seems to be the overturning time scale, at least for the Atlantic, and is what we found in our two-dimensional model.

JONES: Michael, it seems to me that your central England temperature record may show the Little Ice Age, but not the modern minimum. A sunspot feature, perhaps. And Brendan, relative to your Figure 10 with the solar cycle length, I wanted to point out that for that Southern Hemisphere pressure data set you're using there are absolutely no data in the zone between 40° and 50° south. Now, the dotted solar-cycle-length curve shows an enormous cycle about 1750 that Kristen and Lassen have postulated as forcing global temperatures. However, they used that curve from 1850 on while ignoring the record back to 1700.

RASMUSSON: These points are awfully important in critically examining what we really have here. But my question is about the treatment of the tree-ring data. Did you have to remove the growth trend, and if so, how did it affect the low-frequency variations?

JONES: The Tasmanian trees are so much longer-lived than the Scandinavian ones I was referring to that it's less of a problem to remove the growth trend.

BUCKLEY: Our paper does talk about bias, particularly at the end of the series, and describes Ed Cook's technique for dealing with it, which is different from Phil's. Also, yew and pine tend not to follow the normal exponential growth curve; they can be very slow-growing subcanopy trees for 200 or 300 years. Often we don't even use those 300 years because it's hard to measure that small growth with any accuracy.

KEELING: If I may, I should like to show for comparison the analysis we made at Scripps of the Jones-Quigley temperature record. I'm not forgetting, of course, that one tends to favor spectral analyses that confirm one's own position. But it seems to me that those spectral lines do come quite close to those we find in the temperature record. The 4.85 line is not far from one of the principal lines of the El Niño signal. Then there are some that may or may not correspond with others. But that 15-year line, that I mentioned was so hard to explain as solar, seems to be quite persistent. It's also half of the 31-year peak I mentioned in my presentation, which is a principal tidal line. It's not as striking as the 93-year cycle, but it's significant. In any case, whether it's the 80 and 200 you showed, and whether there is a Gleissberg effect or not, wherever you see evenly spaced patterns you should look to see what the beat frequency is, and whether it might relate to a long-period line too.

KUSHNIR: I think we need to remember that if we do a spectral analysis with such high resolution and are looking at the 95 percent significance level, there's always the chance that 5 percent of the peaks will be above the line.

MCWILLIAMS: Dave Keeling's tides have very clear frequencies. But in general, why are we looking for lines in the climate record? Why should we expect a priori to see periodicities rather than more distributive broad-band behavior'?

GHIL: We like periodicities because they enhance predictability. But within the context of non-linear dynamics, even though what you observe is broad-band, that is a reflection of instabilities in the underlying phenomena you are trying to track. My suspicion is that each of those lines has a story behind it, and I hope that over the next 10 or 20 years we can begin to understand the interactions of the various phenomena.

REIFSNYDER: It's also true that some of those lines might disappear as more data are accumulated, though our need for cycles—circadian, budget—won't.

KAROLY: I didn't want to talk about cycles, but mechanisms. I think the mechanism you identified as associated with the pressure and temperature fluctuations in Tasmania suggests a structural change, possibly associated with the Antarctic current or the large-scale atmospheric circulation, that fits in quite well with the modulation of the strength of the zonal winds and the large-scale atmospheric circulation on interannual time scales. And it may well be a mode of the decadal and longer time scales as well, with a broad-band rather than a specific frequency.

REIFSNYDER: We should perhaps keep in mind that statistical significance does not equate with reality.

The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement