By Andy Might
In my final publish, it was recommended that Michael Mann’s 2008 reconstruction (Mann, et al., 2008) was just like Moberg’s 2005 (Moberg, Sonechkin, Holmgren, Datsenko, & Karlen, 2005) and Christiansen’s 2011/2012 reconstructions. The declare was made by a commenter who calls himself “nyolci.” He presents a quote, on this remark, from Christiansen’s co-author: Fredrik Charpentier Ljungqvist:
“Our temperature reconstruction agrees properly with the reconstructions by Moberg et al. (2005) and Mann et al. (2008) with regard to the amplitude of the variability in addition to the timing of heat and chilly durations, apart from the interval c. AD 300–800, regardless of important variations in each knowledge protection and methodology.” (Ljungqvist, 2010).
A fast google search uncovers this quote in a paper by Ljungqvist in 2010 (Ljungqvist, 2010), one yr earlier than the essential reconstruction by Christiansen and Ljungqvist in 2011 (Christiansen & Ljungqvist, 2011) and two years earlier than their 2012 paper (Christiansen & Ljungqvist, 2012). It seems that Ljungqvist’s 2010 reconstruction is kind of totally different than these he did with Christiansen over the subsequent two years. All of the reconstructions are of the Northern Hemisphere. Ljungqvist’s and Christiansen’s are of the extra-tropical (> 30°N) Northern Hemisphere and Moberg’s and Mann’s are purported to be of the entire Northern Hemisphere, however the massive variations lie within the strategies used.
With regard to the realm coated, Moberg solely has one proxy south of 30°N. Mann makes use of extra proxies, however only a few of his Northern Hemisphere proxies are south of 30°N. Determine 1 exhibits all of the reconstructions as anomalies from the 1902-1973 common.
Determine 1. A comparability of all 4 reconstructions. All are smoothed with 50 yr shifting averages apart from the Ljungqvist (2010) reconstruction which is a decadal file. All have been shifted to a standard baseline (1902-1973) to make them simpler to check. Ljungqvist(2010) is a decadal file.
As Determine 1 exhibits, the unique Ljungqvist(2010) file is just like Mann(2008) and Moberg(2005). A few years after publishing Ljungqvist(2010), Ljungqvist collaborated with Bo Christiansen and made the file labeled Christiansen(2012). It begins with the identical proxies as Ljungqvist(2010), however makes use of a special methodology of mixing the proxies right into a temperature file that they name “LOC.”
In 2008, Michael Mann created a number of totally different proxy data, the one plotted in Determine 1 is the Northern Hemisphere EIV Land and Ocean file. EIV stands for “error-in-variables” and is a complete least squares regression methodology. Mann states at first of his paper that he would tackle the criticism (“recommendations”) within the 2006 Nationwide Analysis Council report (Nationwide Analysis Council, 2006). The result’s a fancy and laborious to comply with dialogue of varied statistical strategies used on numerous combos of proxies. He doesn’t have one consequence, however many, then he compares them to 1 one other.
Moberg (2005) additionally makes use of regression to mix his proxies however characterizes them by decision to protect extra short-term variability. The statistical approach utilized by Ljungqvist in his 2010 paper is analogous and known as “composite-plus-scale” or CPS. This system can also be mentioned by Mann in his 2008 paper and he discovered that it produced related outcomes to his EIV approach. Since these three data had been created utilizing related strategies, all of them agree fairly properly.
Christiansen and Ljungqvist (2011 and 2012)
Everybody admits that utilizing regression-type strategies to mix a number of proxies into one temperature reconstruction reduces and dampens the temporal decision of the ensuing file. Instrumental (thermometer) measurements are usually precisely dated, at the very least all the way down to a day or two. Proxy dates are a lot much less correct, a lot of them are usually not even recognized to the yr. These which can be correct to a yr typically solely mirror the temperature through the rising season, throughout winter or through the flood season. Ljungqvist’s 2010 file is barely decadal resulting from these issues.
Inaccurate dates, irrespective of how rigorously they’re dealt with, result in mismatches when combining proxy data and end in unintentional smoothing and dampening of high-frequency variability. The regression course of itself, results in low-frequency variability, Christiansen and Ljungqvist write:
“[Their] reconstruction is carried out with a novel methodology designed to keep away from the underestimation of low-frequency variability that has been a common downside for regression-based reconstruction strategies.”
Christiansen and Ljungqvist commit lots of their paper to explaining how regression-based proxy reconstructions, just like the three proven in Determine 1, underestimate low-frequency variability by 20% to 50%. They record many papers that debate this downside. These reconstructions can’t be used to check present warming to the pre-industrial period. The century-scale element, previous to 1850, merely isn’t there after regression is used. Regression reduces statistical error, however on the expense of blurring essential particulars. Due to this fact Mann including instrumental temperatures onto his file in Determine 1 is mindless. You may as properly splice a satellite tv for pc photograph onto a six-year-old little one’s hand-drawn map of a city.
Christiansen and Ljungqvist ensure that all their proxies have a great correlation to the native instrumental temperatures. About half their proxies have annual samples and half decadal. The proxies that correlate properly with native (to the proxy) temperatures are then regressed in opposition to the native instrumental temperature file. That’s the native temperature is the impartial variable or the “measurements.” The subsequent step is to easily common the native reconstructed temperatures to get the extratropical Northern Hemisphere imply. Thus, solely minimal and mandatory regression is used, in order to not blur the ensuing reconstruction.
Regression does cut back the statistical error within the predicted variable, however it reduces variability considerably, as much as 50%. So, utilizing regression to construct a proxy temperature file to “show” current instrumental measured warming is anomalous is disingenuous. The smoothed regression-based data in Determine 1 present Medieval Heat Interval (MWP) to Little Ice Age (LIA) cooling of about 0.8°C, that is extra more likely to be 1°C to 1.6°C, or extra, after correcting for the smoothing resulting from regression. There may be further high-frequency smoothing, or dampening, of the reconstruction resulting from poorly dated proxies.
The extra cleverly constructed Christiansen and Ljungqvist file (smoothed) exhibits a 1.7°C change, which is extra in keeping with historic data, borehole temperature knowledge, and glacial advance and retreat knowledge. See Quickly and colleagues 2005 paper for a dialogue of the proof (Quickly, Baliunas, Idso, Idso, & Legates, 2003b). Christiansen and Ljungqvist keep a lot nearer to the information of their evaluation to keep away from distorting it, this makes it simpler to interpret. Determine 2 exhibits the identical Christiansen and Ljungqvist 2012 curve proven in black in Determine 1 and the yearly Northern Hemisphere averages.
Determine 2. Christiansen and Ljungqvist 2012 50-year smoothed reconstruction and the one-year reconstruction. The black line is similar as in Determine 1, however the scale is totally different.
The one-year reconstruction is the high quality grey line in Determine 2. It’s a easy common of Northern Hemisphere values and is unaffected by regression, thus it’s as near the information as doable. Most variability is retained. Discover how briskly temperatures differ from year-to-year, generally by over two levels in only one yr, 542AD is an instance. From 976AD to 990AD temperatures rose 1.6°C. These are proxies and never exactly dated, so they aren’t actual values and take them with a grain of salt, however they do present us what the information say, as a result of they’re minimally processed averages.
The complete vary of yearly common temperatures over the two,000 years proven is 4.5°C. The complete vary of values with the 50-year smoothing is 1.7°C. Given that almost half of the proxies used are decadal, and linearly interpolated to one-year, I belief the 50-year smoothed file greater than the yearly file over the long-term. However, seeing the variability within the one-year file is illuminating, it reinforces the foolishness of evaluating fashionable yearly knowledge to historic proxies. Fashionable statistical strategies and computer systems are helpful, however generally they take us too distant from the information and result in misinterpretations. I believe that always occurs with paleo-temperature reconstructions. Maybe with fashionable temperature data as properly.
It’s fairly doable that we’ll by no means know if previous climatic warming occasions had been quicker than the present warming charge or not. The high-quality knowledge wanted doesn’t exist. What we do know for positive, is that regression strategies, all regression strategies, considerably cut back low-frequency variability. Mixing proxies with various resolutions and imprecise dates, utilizing regression, destroys high-frequency variability. Evaluating a proxy file to the fashionable instrumental file tells us nothing. Determine 1 exhibits how vital the statistical strategies used are, they’re the important thing distinction in these data. All of them have entry to the identical knowledge.
Obtain the bibliography right here.