By Andy May
I’m reviewing the Pages 2k temperature proxies one-by-one as I once reviewed Marcott’s proxies (May, 2018). There are 647 Pages 2k proxies, so this is a big job. This reconstruction is prominently displayed and used in the IPCC AR6 report (IPCC, 2021, pp. 211, 432) so it is worth the time to investigate it. Their proxies can be downloaded as one R data file (RData format) from here. This post will just cover six of the Greenland ice core proxies, but the full ensemble median Pages 2k global temperature reconstruction is shown in figure 1.
Figure 1 resembles the well-known “hockey stick” reconstruction (Mann et al., 1998), which has been the subject of substantial methodological criticism regarding proxy selection, principal-component methodology, and uncertainty estimation. It was featured prominently in the third report or “TAR” (IPCC, 2001). We will not get into the problems with the previous hockey stick here, they are well covered by Steve McIntyre (McIntyre, 2011), (McIntyre & McKitrick, 2005), (Wegman et al., 2010), and Andrew Montford (Montford, 2010).
Pages 2k Statistical Methods
Some of the problems with figure 1 are statistical and similar to the problems with Mann’s hockey stick. This post is about selected and well-known Greenland ice core proxies and the technical problems using them to compute temperatures over the past 2,000 years, but it is worthwhile to briefly list the statistical problems and provide references for more details.
The Pages 2k consortium uses seven different temperature reconstruction methods, including principal component regression, arguing that by using multiple methods and getting broadly similar results, confidence in the outcome is increased (Pages2K, 2019). However, if all seven methods are fed the same data and calibrated to the same instrumental data from ~1850 to 2014, agreement among the methods is not independent evidence nor surprising. The real problem is the calibration period is short, only about 150 to 170 years and it lies at the end of the proxy records. This is the same problem Mann had, calibration points at only one end assume that proxy sensitivity to temperature has not changed over the past 2,000 years, which is unlikely.
In addition, 392 of the 647 proxies are tree-ring proxies. I’ll deal with the problems with using tree rings as temperature proxies in another post, but for now just remember that tree ring records are detrended, which “results in a loss of multi-century variance” (Briffa et al., 1998b). This is the “segment-length curse” (Cook et al., 1995), which means that recoverable climatic information can be limited to the lengths of the individual tree-ring series used to build a reconstruction. The Pages 2k team tries to get around this by using “Regional Curve Standardization” (Esper et al., 2003), but in order for this to work properly all trees in the chosen standardization region must belong to the same “biological growth” population, something that cannot be verified 2,000 years ago, or even several hundred years ago. “This problem still remains largely unsolved.” (Cook et al., 1995)
The Pages 2k database paper (Pages 2K, 2017) admits that less than half of the proxies they studied correlated with HadCRUT4.2 over the instrumental period (1850-2014). Proxy records that did not significantly correlate with instrumental era measurements were rejected (Pages 2K, 2017) & (Pages2k, 2013). Screening proxies based on correlations during the short modern calibration interval may preferentially retain proxies that share characteristics of the late twentieth-century climate trend, potentially introducing selection bias. This selection bias can have the effect of forcing a hockey stick shape (Wegman et al., 2010). Further, the use of principal components, and some other dimensionality reduction techniques can cause similar problems. Principal component methods reduce many proxy series into a smaller number of composite indices. Depending on preprocessing choices and network composition, these methods may emphasize certain patterns of variability (such as modern warming) while suppressing others.
Pages 2k is attempting to produce a global mean temperature, but their spatial proxy coverage is uneven, even sparse in the Southern Hemisphere. Overall, their methodology is overly complicated and as methodological complexity increases, transparency decreases.
Pages 2k is not a simple repeat of Mann 1998, but its use of multiple statistical methods does not automatically eliminate shared biases, especially if all non-correlating proxies in the very short calibration period (~1850-2014) are rejected. If all methods use the same proxy network, the same calibration period, and similar assumptions about proxy stationarity, then agreement among methods is not proof that the reconstruction is correct. “Proxy stationarity” is the assumption that the relationship between the proxy and the target variable observed in the calibration period remains constant over time. For a good discussion of some of the problems with Pages 2k, see Steve McIntyre’s South American proxies, North American tree ring proxies, and Antarctic proxies.
Greenland
The main goal of this post is to help the reader appreciate why proxy temperatures are uncertain, even though ice core proxies are probably the best temperature proxies. A “proxy” can be a singleton, for example a single ice core, or a small region “reconstruction” or “aggregate” based on multiple cores from the same area that have been carefully correlated. Thus, the term is somewhat vague. In this post we will use “proxy” as synonymous with a small region reconstruction, since many ice core records are constructed from multiple cores taken in a small area.
The Pages 2k dataset contains six Vinther proxies (Vinther et al., 2009) from the Agassiz area in northeastern Canada and Greenland that we will focus on here. They are mapped in figure 2 and all were used in the Pages 2k final global reconstruction.
The Vinther Greenland proxies are well studied and have been used to make a Greenland temperature reconstruction that, in my opinion, is the best. This is the Vinther 2009 (Vinther et al., 2009) reconstruction shown alongside the Rosenthal Makassar Strait reconstruction in figure 3. In this post we will treat the Vinther 2009 reconstruction as the ideal for Greenland.
The two reconstructions shown in figure 3, Vinther’s 2009 Greenland reconstruction (the orange line) and Rosenthal’s 500-meter depth Makassar Strait reconstruction (Rosenthal et al., 2013) of the North Pacific SSTs (black line) are high-quality and have high temporal resolution (~20 years). They are also 9,500 miles apart.
Vinther’s 2009 Greenland Reconstruction
In 2009, Bo Vinther published a now-classic paper that explains a lot of problems with Greenland ice-core proxies and his solution for some of them. The oxygen-18 (“18O”) to oxygen-16 (“16O”) ratio from ice cores used in climate studies is a comparison of the measured ratio to the VSMOW2 standard ratio in ‰ (per thousand) difference. The standard is conceptually based on the mean seawater 18O /16O ratio. In reality, VSMOW2 is comprised of a mixture of lake water samples and is formulated to match the original seawater standard (SMOW) developed by Harmon Craig in the 1960s (Craig, 1961).
Vinther recognized that the 18O /16O ratio (or more simply, δ18O) was not simply a function of temperature, but also a function of the elevation of snow deposition, which varies substantially over time as a result of tectonic changes and the rate of snow deposition and glacier melting or movement. In figure 2 the current elevation is posted at each proxy site. Vinther recognized that the δ18O records from the Agassiz and Renland sites are similar, and dissimilar to the other sites. The four Agassiz cores are compared to the Renland core and the NCDC high resolution versions of the proxy records in figure 4.
The Agassiz A77 core is over two kilometers from the other three Agassiz cores and down slope from them, as a result its δ18O record is offset 1-1.5 ‰ negative as shown in the left graph in figure 4. This is likely due to scouring of winter snow off the Agassiz dome and ridge, where the other three cores are located, and the deposition of excess winter snow on the A77 site (Fisher et al., 1995) & (Fisher et al., 1983). This causes the mean annual δ18O to be lower at A77, a systematic seasonal effect. The plot of the offsets in the right-hand graph supports this idea, the offsets are similar and they became more negative in the Little Ice Age when the winters were colder, that is the summer/winter difference was more extreme in the Little Ice Age. This was a factor in the increased storminess observed in the Little Ice Age (Behringer, 2010), (Parker, 2012) & (Costas et al., 2016).
Unlike the ice cores taken deep into the Greenland ice sheet, the smaller Agassiz and Renland icecaps have reasonably accurate elevation histories available and their δ18O records can be corrected. Once this was done, they were very similar even though the sites are 1,500 km apart and they are on opposite sides of the Greenland Ice Sheet (Vinther et al., 2009). Vinther’s corrected δ18O Greenland temperature reconstruction is shown in figure 3 with the orange line.
Converting δ18O to ΔT
Creating an accurate and defensible temperature anomaly (ΔT) from ice core δ18O is challenging. The list of things that affect ice core δ18O besides temperature is long, it includes elevation, wind-blown snow, topography, and location. There are two methods, one is to correlate δ18O to local instrumental temperatures when they are available and fit an ordinary least squares (OLS) or other linear best line to the data and use the resulting slope to compute ΔT for the whole record. In Greenland this methodology produces very poor regression statistics as shown in figure 5 for the NCDC/Pages 2k one-year Agassiz δ18O record (R2 = 0, p=0.92) and the 5-year Renland record (R2 = 0.267, p = 0.01).
Figure 5 is constructed from Pages 2k δ18O high-resolution data and local NOAA GlobalTemp version 6. The correlations are very poor. It doesn’t get any better using the original Vinther 20-year sample version of the same ice core data. GlobalTemp data from the closest grid cell is used on the y axis and the site δ18O is on the x axis. These correlations are not unique, nearly all the Greenland Pages 2K proxies produce poor R2 values as shown in table 1.
Besides the poor R2 and p values in table 1, the slopes are odd, especially considering that Agassiz A87 (slope = -20 ‰/°C) and Agassiz A84 (slope = 0.369 ‰/°C) are only about 30 meters apart and A79 (slope = 0.8 ‰/°C ) is only a km away. These regressions do not provide a robust basis for estimating temperature sensitivity. We used the fits to compute ΔT and the result is plotted in the right-hand graphs in figure 5. The Agassiz plot is essentially a horizontal line, and the Renland plot is close to a horizontal line. Pages 2k explains that extratropical winter variability is a better predictor of the annual average, so when given a choice (as they are here) they use winter values (Pages 2K, 2017). The proxies used by Pages 2k, in table 1, have an “NCDC” prefix.
The Pages 2k documentation (Pages 2K, 2017) indicates that instrumental correlations played a role in calibration of the proxy to temperature relationship and in screening the proxies. I was unable to locate sufficient information to fully reproduce their calibration process. They do not supply the calibration datasets or the proxy-by-proxy correlation statistics that I could find and the explanation they do provide (Pages 2K, 2017) suggests the process is quite complex. In my work, I used all points between 1870 and 2020 to do the calibrations.
NOAA GlobalTemp
The NOAA GlobalTemp instrumental record for the grid cell identified in figure 2 is compared to the SW Greenland instrumental record (Vinther et al., 2006) in figure 6.
Figure 6 compares the land only SW Greenland temperature record and the corresponding record for the closest NOAA GlobalTemp grid cell located in figure 2. The grid cell contains a lot of sea surface area, that have temperatures that warm more slowly than the land-based temperatures in the SW Greenland record, which probably accounts for the 20th century difference.
Using “standard” δ18O to local temperature conversions
The problem with using an ice-core proxy δ18O to local temperature correlation for predicting temperature was recognized in the 1960s and Dansgaard and colleagues developed local correlations between δ18O in precipitation and temperature where both measurements are from the same sample. Their data is plotted in figure 7. Isotopic fractionation of precipitation versus condensation temperature is a real thing and atmospheric δ18O does decrease with condensation temperature. However, it is not a simple correlation and there are many confounding local factors when comparing regional temperatures to δ18O proxies. These include wind speed, relative humidity, topography, storm tracks, sea ice extent, elevation, and seasonality of precipitation. Thus, using a conversion from δ18O to ΔT varies from location to location.
Below is a blow up of Greenland points away from the coast shown in green in figure 7.
Figure 8 shows just the points in the interior of Greenland. All temperatures and δ18O value are measured from the same samples in the same conditions. The correlation is excellent and the resulting slope of 0.63 ‰/°C is very close to the generally accepted Greenland conversion of 0.62 ‰/°C used for the curves in figure 9.
Conclusions
As should be clear from figure 7, the slope of 0.62 ‰/°C is only appropriate for ice cores in the interior of Greenland. It is not appropriate for other ice cores, marine carbonates, foraminifera, corals, speleothems, or lake carbonates. In fact, from my OLS study of the Pages 2k δ18O proxies the average derived slope for 46 marine and lake sediment samples was -0.77 ‰/°C. All these OLS slopes are highly speculative with an average R2 of 0.16. The normal marine slope is negative and found by linear regression, no dataset like shown in figures 7 & 8 exists for marine sediments to my knowledge. These results suggest that temperature reconstructions or aggregates based solely on marine δ18O should be interpreted cautiously because non-temperature influences can be large and difficult to quantify. The Rosenthal reconstruction plotted figure 3 is a bit unique, since it is very well located, with an accurate depth that is corrected for sea level changes, and very well constrained (Rosenthal et al., 2013).
When a good correlation between δ18O and ΔT is available, and the relationship can be reasonably expected to persist into the past over the time span of the proxy record, δ18O can be a very good proxy. I think that is the case for Greenland ice cores from the interior (see figure 8), as long as elevation changes can be corrected for and the sites are well located. Agassiz 77 is a good example of a poorly located proxy.
Site specific problems
While δ18O in precipitation is differentiated by condensation temperature, there is no unique one-to-one relationship between proxy δ18O and temperature because multiple climatic and environmental variables influence the isotopic signal. In a nutshell, water molecules containing 18O are slightly heavier than those containing 16O. As moist air cools and condenses, the heavier 18O preferentially rains or snows out. As an air mass moves poleward or upslope, it becomes progressively depleted in 18O. Colder condensation temperatures therefore generally produce more negative δ18O values. This produces the familiar positive slope observed in figure 7.
Corals, foraminifera and other marine carbonates are different. Colder seawater causes carbonates to incorporate more 18O. Thus, colder water has a higher δ18O and the slope becomes negative, we normally expect a slope of about -0.2, but the Pages 2k average is much higher than this, so these slopes are highly uncertain. Ocean water 18O content is affected by evaporation, precipitation, river input, sea ice formation, and melting ice, so the 18O content in shells and corals is very hard to interpret. The relationship between δ18O and ΔT, everything else equal, is linear but the confounding factors of humidity, moisture source, storm tracks, seasonality, evaporation/precipitation balance, ice volume and proximity, and salinity make the relationship quite uncertain.
Transparent calibration procedures
The importance of the calibration of δ18O to ΔT is apparent. This makes the lack of specific proxy-by-proxy details in the Pages 2k documentation and papers unsettling. All I could find was the phrase “correlation with instrumental” deep in the RData documentation.[1] No statistics were stored in the record, and the conversion process is only vaguely described in the papers and other documentation (Pages 2K, 2017). Transparent calibration proxy-by-proxy details are critical because inferred temperatures depend strongly on the assumed δ18O-to-temperature conversion, and the conversion depends upon local conditions.
Combining multiple proxies into one global or hemispheric reconstruction is unlikely to work and hard to interpret. Each proxy is influenced by local conditions that are not necessarily related to temperature. All proxies are not equally accurate, especially over long periods of time, and the long-term accuracy is difficult to estimate. The correlation to modern instrumental temperature measurements may not persist into the past, and we could not verify that it does even if it did. It is easier to interpret proxies one at a time, each has unique problems, as we can see with the Greenland examples in this post. The only meaningful use of temperature proxies is to estimate the local temperature history around the proxy or aggregate proxy location, which should be carefully compared to local historical or archeological records (Soon & Baliunas, 2003) & (Soon et al., 2003b).
The R code to make these figures and perform the analysis is complicated so I will not provide a code supplement in this post, instead I will write a post or two describing the code and provide downloadable code and data with those posts.
Download the bibliography here.
Under: site name$paleodata$measurementTable$d18O$interpretation ↑
Very interesting post. The comparison between Figure 1 and Figure 3 is striking.
Yes, y-axis range on F3 is 5.0C … y-axis range on F1 is 1.5C.
I don’t understand why people who make misleading charts don’t understand that anyone who actually looks at their chart will immediately notice what’s misleading.
Andy, how do they calculate the ridiculous uncertainty numbers in Fig. 1?
Good question. They use a bootstrap approach, which is a fancy way of saying they are measuring the sensitivity of the reconstruction to which proxies are included. It is done by repeatedly changing the proxies used (with replacement) and measuring the differences. This technique explicitly does not measure proxy calibration uncertainty, dating uncertainty, it does not check stationarity, the uncertainty due to the statistical methods used, or any bias uncertainty. Yet all these problems exist.
The method essentially says my measurements are OK, the error is in which ones I use.
For more details see here: Pages 2K. (2017). Data Descriptor: A global multiproxy database for temperature reconstructions of the Common Era. Nature, 4. https://doi.org/10.1038/sdata.2017.88
Aha thanks. I wrote a comment about the strange uncertainty that you’ve explained here. Good article, offers lots to think about.
One more key point. Bootstrap intervals are generally very good for estimating sampling variability, but they can understate total uncertainty when systematic errors dominate. That distinction is often important in large multiproxy reconstructions. In this case, systematic errors dominate, I think I showed that very clearly in this post.
Quite a few years ago, I posted an active viewer for the Pages2k proxies here. No doubt there is new information since, but I found it a good way of viewing a lot of information.
Thanks Nick. Am I right that it just shows the proxy in the proxy units? It doesn’t show the converted temperature as far as I can see. I suspect most of the uncertainty is in the conversion, which was the point of my post.
I think most of the measurements are OK, its just that the conversion takes a hands on approach since every proxy is unique and has its own unique problems. This is one reason I like Vinther and Rosenthal’s work so much.
Yes, Andy, it does not include the conversion, but gives a reference to the original paper.
Andy,
Praise for taking on this much-needed, lengthy, dreary task of calling out the serious problems of PAGES2K. I have followed this since the start, with its Australian content.
Please let me know if I can help, even with basic aid like digging up references that are hard to find.
It is big enough to consider organising a citizen science support group. Poor science like PAGES2K has to be exposed because it is harming the public view of Science, which is not helpful.
Geoff S
Thanks Geoff. I’ll let you know if I need help.
“Less than half” it would be interesting to know more precisely what percentage of samples were used to invent these ‘hockey stick’ reconstructions.
That’s an understatement. The so-called ‘calibration period’ process is circular reasoning it is where the chicanery starts but sounds so plausible, randomly selected samples of course would not yield the desired result.
Around the globe, the climate changes. Antarctica used to support thriving fauna and fauna – now it’s a frozen desert.
The Sahara was lush, green, and fertile – now it’s a hot desert.
All this fake “research” (like Pages 2K) is the product of fools or frauds, designed to convince the ignorant and gullible that adding CO2 to air makes thermometers hotter. It doesn’t, of course.
The Earth’s surface was once everywhere molten – now it’s not.
That’s called cooling – four and a half billion years of it. When a bit of the surface rises, another bit has to fall. When a portion of the surface gets hotter, another portion is deprived of heat – and cools.
The branch of science trying to find the reasons is called physics.<g>
Any mention of Vikings, crocodiles by IPCC?
No, but plenty of pigs feeding at the trough.
The uncertainty shade region on the Pages 2k reconstruction is not symmetrical, meaning possible “plus” error is a little different than possible “minus” error. I don’t know what it means, but I do know it took extra effort not to apply boring symmetrical three-sigma over/under the mean. I’d applaud the chart-maker for going an extra – then I think, “I wonder whether they were running from a huge error range that would have made their chart silly if they took the easy way?”
The error shown in figure 1 is only the error possible from the proxies selected, that is bootstrap error. The major sources of error, in reality, are bias error, conversion to temperature (discussed in this post), dating uncertainty, stationarity (how far back does the calibration to instrumental go?), and the uncertainty due to the statistical methods used. If they were able to measure these effects, all of which apply to this reconstruction, and present them, the error would be close to infinite. That means the answer to your question:
Is yes.
A good look as the possible error (unmeasured in the uncertainty in figure 1) is given in figure 4. A77 is only about 2 km from the other three Agassiz ice cores, but it is systematically about 2-3 permill lower, which is about 4 deg C colder over 2000 years. 0.62 ‰/deg inverted is 1.61 deg/‰, so 2.5×1.61=4 deg. Compare that to figure 1 and get an idea how bad it truely is.
Hi Andy, good post, an eye-opener on the uncertainty of PAGES2K analyses and the cherry-picking that’s going on. What struck me most however was the synchronicity between the Vinther and Markasser straights data that shouldn’t have been there: Greenland climate change should’t be the covary with equatorial climate change. The last glacial maximum occurred 23 ka ago, when Earth’s perihelion was during the December solstice. The Greenland Holocene Thermal Optimum in your figure 3 occurred around 9000 BC, when Earth’s perihelion was around the June Solstice (9564 BC), and solar insolation at 65 ºN at a maximum. Greenland temperatures declined again towards present, towards the 1246 AD December solstice perihelion. All of this is as expected. But why did the Markassar straits follow suit? The seasons would have suffered an equinox flip. Around 9564 BC, the equator received an intense spike of solar radiation during the northern summer months. Today, that seasonal spike has shifted to the southern summer months. The total energy would have remained roughly constant. It would appear that the North Atlantic is controlling global temperatures. Also: on sub-orbital time-scales it appears your figure 3 might have some Eddy (1000 year) cyclicity happening. All of these might be good clues as to what is currently happening in terms of natural climate change variability.
The best explanation is given in Rosenthal’s 2013 paper in the bibliography, but I’ll summarize. The fossils and shells studied by Rosenthal and his colleagues are from a bottom-dwelling foraminifer that lives at about 500 meters depth in the Makassar Strait. Sea level varied over his study period, but he has corrected for this. Five hundred meters is deep enough to be insulated from short-term weather fluctuations on the chaotic surface, but shallow enough to reflect longer-term surface climatic fluctuations. More here:
https://andymaypetrophysicist.com/2021/06/23/how-to-compare-today-to-the-past/
The beauty of the proxy is that it is not a sea level proxy, but deep into the slower changing mixed layer. Today it tracks the surface temperature over most of the North Pacific, but is also influenced part of the year by the Indian and Southern Oceans. In other words it is a nearly perfect location.
“Proxy records that did not significantly correlate with instrumental era measurements were rejected”.
Expletive deleted.
That gets me every time I read it. I’m no trained statistician but I’m sure I would have been thrown out of my Master’s program had I only selected the analytical results that “significantly correlated” with my model and rejected all those that didn’t.
The quote is that the proxies that didn’t match the instrumental record were rejected. It said nothing about models.
Lol…that’s not what PAGES2K did.
From the head post:
“The Pages 2k database paper (Pages 2K, 2017) admits that less than half of the proxies they studied correlated with HadCRUT4.2 over the instrumental period (1850-2014). Proxy records that did not significantly correlate with instrumental era measurements were rejected (Pages 2K, 2017) & (Pages2k, 2013).”
Sure sounds like what PAGES2K did …
w.
Stop gaslighting Willis. They did NOT “selected the analytical results that “significantly correlated” with my model and rejected all those that didn’t.” They did the opposite of that. They rejected the proxies that did not significantly correlate with observations. Any half-brained scientists in any discipline would consider that acceptable. And to not do that could be considered unethical at best and fraudulent at worst depending on the circumstances.
bdgwx,
Not in this context.
None of the proxies correlate well with observations.Somehow they drew a line and said above this line we keep, but below this line we reject. They do not explain where that line is.This risks selection bias.If a class of proxies had R^2 of >0.9, and a few were much less, say 0.2, then what you say is true, but R^2=0.9 is very rare in this dataset, the average d18O r^2 is 0.16, in technical terms, that is crap. What did they do, accept with an r^2 >0.08?
The proper way to do it, in my opinion, is examine each proxy individually, in the context of its location and historical records, then accept it if it “looks right” but only use it by itself, don’t statisticaly combine it with other proxies. This data is not good enough to do that. Each proxy is unique and likely to have unique problems due to its location and type. Combining proxies is a fools errand. I should know I spend a lot of time trying to do just that before I realized it was foolish.
https://andymaypetrophysicist.com/2017/06/09/a-holocene-temperature-reconstruction-part-4-the-global-reconstruction/
Thanks, bdgwx.
I’m sorry, but selecting temperature proxies based on their correlation with modern temperature records and discarding proxies without correlation results in a hockeystick. Every. Single. Time. That procedure actively mines for a hockeystick, and produces one out of red noise.
And appealing to “half-brained scientists” doesn’t change that at all. Try it yourself. Generate 1,000 red-noise pseudoproxies covering the last say 1,000 years, and select only the ones that correlate with modern temperature. Average them, and you’ll get a hockeystick.
That’s the problem.
w.
You are correct Willis, it is exactly what they did. It was a critical decision and they bury it in their overly complicated process, but they did it. And if they documented the criteria, I couldn’t find the documentation. Just vaguaries. Here is the key bit from their documentation:
At least they admit the problems with the decision.
The other criteria they use are OK, not enough points in the calibration interval (1850-2014), poor dating, etc. These are very objective. They also screen using nearest neighbor, but sometimes that is thousand of km away, so that is problematic. Just look at the Agassiz A77 in the post as an example.
Stop gaslighting Andy. They did not reject the observational record in favor of the model. That’s what Phil R said. And I don’t mean he insinuated it…he literally said it…and I quote…“I’m no trained statistician but I’m sure I would have been thrown out of my Master’s program had I only selected the analytical results that “significantly correlated” with my model and rejected all those that didn’t.”
And rejecting a model (in this case the proxy) because it does not significantly correlate with the observations will not get you kicked out of college. In fact, unless there is some other valid justification in play then to not reject a model (in this case the proxy) because it does not correlate with observations could be considered unethical at best and fraudulent at worst. Don’t ever let it be said that I defend Phil R’s position here.
bdgwx,
Give it up, the whole crazy idea of selecting proxies based on their end points has been shown to cause hockey sticks out of red noise many times, it always does. Below Willis shows it, this is one of three times I’ve seen this proven in various ways, the first two were Wegman and McIntyre, both are either linked or in the bibliography. This is not speculation, it is proven.
You are deflecting and diverting. That isn’t the topic of this subthread. The topic of this subthread is the indictment of wrong doing because a model was rejected because it didn’t match observations. Mike Jonas and Phil R’s position is indefensible because it is often appropriate and, in fact, to not do it can be unethical or even fraudulent. I’ll give Willis some grace here because I’m not sure he fully understood what he was defending.
“That isn’t the topic of this subthread.”
Huh? That is exactly the subject of this subthread. I don’t think Willis is confused, I think you are. Everyone agrees that overall temperatures warmed in the 20th century, there is considerable disagreement on how unusual it is, but it did warm. Here we have a set of over 600 proxies that, with very few exceptions, do not significantly correlate with that warming. p values well over 0.1. There is no valid basis for accepting or rejecting any of them based on correlation statistics, especially when it is well known that selecting them based on 20th century correlations causes a hockey stick, even with red noise.
I had the same thought. Why reject any proxy? There should be a reason they didn’t correlate other than “they didn’t correlate”! Unless well explained one could imagine that it would have skewed the results in an unwanted direction.
Thanks for the response. in my academic career (such as it was; Master’s in geology) and in my professional career, whenever we had to deal with analytical data we were always told not to throw any out and present the total data, warts and all. Then if we did any analysis where data was left out, such as extreme outliers, etc., that was ok but we had to specifically and directly note what was left out and why. “Did not significantly correlate” with what I expected or the client wanted was not a valid reason.
They are looking for temperature proxies. If the proxies don’t correlate with temperature in the modern era, why should they assume the proxies would correlate with temperatures in the past.
None of the proxies correlate with the modern temperature record very well over the industrial era, so where do you draw the line? Why reject tree ring series A and keep tree ring series B? They don’t explain this. IMHO they should examine each series, with its warts and all, and make a decision based on each study why to keep it or reject it, then document the decision. Using an arbitrary statistical cutoff just risks introducing selection bias.
The real problem, that Pages 2k completely overlooked in their study, is that none of the proxies correlate well with measured temperatures from 1870 to 2020. They simply rejected those that failed the statistical test by a wider margin. Had they given it more serious thought, they would have realized that their automated process was crap and would never work. A proper study of temperatures in the past 2,000 years means studying historical records first, then carefully choosing the very best located, preserved, and undisturbed proxies (location by location) that match history. Then deal with them one by one. They have turned the natural, and more scientific, procedure on its head. Statistics does not replace common sense.
The common sense solution only leads to high quality reconstructed records in the best locations, it does not result in a single “global” record, which is why common sense was abandoned.
My cynical and puerile Saturday morning comment is, the common sense solution only leads to high quality reconstructed records in the best locations, it does not result in publications, free vacations and grant money.
Would you mind posting a link to historical records prior to 1850 that is robust enough to do what you are suggesting? I would like to review it if you don’t mind.
Sure, Start here:
https://andymaypetrophysicist.com/2015/11/26/climate-and-human-civilization-over-the-last-18000-years-updated/
Behringer’s book in the bibliography is very good. So is Geoffery Parker’s book. Finally do not bypass the 2003 papers by Willie Soon and Sallie Baliunas. All these references are in the bibliography, the Soon and Baliunas papers provide the most concise lists of climate science relevant events.
I don’t see any link to a robust historical record prior to 1850 at your link. And I’d be surprised if there were any considering thermometers were not widely used to measure air temperature until around 1850. Even the invention of a thermometer didn’t occur until the 1600s. So taking a “historical record” first approach is going to be extremely problematic at best since they wouldn’t be anchored with actual temperatures.
The very best indicators of climate are glaciers (this has been known since Tyndall in the 19th century), their advances and retreats are part of the geologic record, however other sensitive and accurate indicators are pollen records, date of ice cover over lakes and rivers. These and other records are mainstream geology. See here for an intro and bibliography: https://andymaypetrophysicist.com/2026/02/13/holocene-glacier-records/
Ignoring common sense climate indicators and accepting dubious proxy records is foolish. They must be used together.
BTW…I’m not suggesting that it is wrong to keep proxy records that do not correlate with the observational record especially if there are other robust way to confirm their legitimacy. What I’m pushing back on is Mike Jones’, Phil R’s and Willis’ indignation that it is wrong to reject them because of the lack of correlation. Though in Willis’ case I’m not sure he understood that Mike Jones’ and Phil R seem to have misunderstand what PAGES2K actually did and perhaps Willis’ thought I was saying you should always keep proxy records that do not correlate with the observational records which I’m obviously not defending either. I don’t know. I’m just saying that the insinuation from this subthread that “Proxy records that did not significantly correlate with instrumental era measurements were rejected” is wrong and that it would get you kicked out of college is itself wrong.
I think all we are saying is that rejecting a proxy just because it does not correlate at some level the very end of its data is wrong. Remember these are 2000 year proxies, sometimes with 20 year sampling intervals or worse. The whole instrumental period is only about 170 years at best. You can reject records with good reasons, but document them!
And, in my opinion, a poor correlation with HadCRUT4, when nearly all the proxies have poor correlations is not sufficient reason. Look into the history at the proxy site first. the correlation is not sufficient reason in this dataset. In my opinion, only a few proxies have good statistics in the calibration interval, I don’t know how many offhand have r^2 over 0.9, but it is just a handful.
If the proxy doesn’t correlate to temperatures over the time period where we have good temperature records, then the proxy is not a temperature proxy and shouldn’t be used to reconstruct ancient temperatures.
One of the complaints regarding the use of tree rings as temperature proxies, is that they do not correlate to temperatures in the modern era.
Your point is very valid. But, it is also true that if you have a thousand tree ring records and you only keep those that fit 1870 to 2020, meaning you reject over half of them, you introduce a possible selection bias. I would say it is OK to go through them one-by-one and examine each and the correspondence to history and archeology in the region where the trees are and then use them to draw conclusions about the specific location. But that is not what Pages 2k did, thus the risk of introducing selection bias.
They tried to automate the process, while simultaneously ignoring the possiblity of systematic and conversion errors and bias.
“They tried to automate the process, while simultaneously ignoring the possiblity of systematic and conversion errors and bias.”
I don’t think they ignored the possibility, I think they relied on it.
Is that proxy being correlated to a “global” temperature? Or to a site that is as close as possible to the proxy location? If the former, how is that meaningful in any way?
Andy,
Eldrosian held up this paper,
High Arctic Holocene temperature record from the Agassiz ice cap and Greenland ice sheet evolution | PNAS
as evidence of AGW in that, and I quote, “Modern temperatures are about 4°C above their preindustrial baseline and are now the highest in roughly 7,000 years.”, and that the rate of change is unprecedented in more than 11,000 years. I had doubts that either claim was well supported by the presented data, but I lack your skill in dissecting the methods and interpretation.
Have you evaluated it?
I determined that, as you say, Agassiz is a poor proxy location.
I have spend a lot of time (too much probably) on this paper. In summary these are my thoughts:
The paper makes some interesting points, especially their speculation on the migration of the warming across Greenland. But their Huy3 model is flawed, which invalidates their conclusions.
Sites like Agassiz, near the Greenland Ice Sheet margins, may overestimate HCO (Holocene Climatic Optimum) warmth due to elevation-driven cooling during ice buildup or warming during retreat. For instance, Kaden Martin et al.’s 2024 study on Greenland ice cores notes that marginal records can be “overprinted” by such effects, leading to discrepancies with central sites like GISP2 and NEEM.
Lecavalier’s corrections (using lapse rates of ~7°C/km) are applied, but later work suggests these may underestimate variability, implying the HTM temperature peak could be less extreme or differently timed regionally. Similarly, a 2018 (Strunk, 2018, doi: 10.3389/feart.2018.00129) study on North Greenland RSL argues that the Huy3 model (co-developed by Lecavalier) underestimates LGM ice thickness, suggesting the GrIS was thicker and more extensive, which could alter Holocene temperature interpretations.
Thus, while an interesting paper, Lecavalier’s conclusions are very suspect. Look closely at figure 9 in Strunk et al. The range of elevations from 12,000 yrs ago to 2,000 years ago is over 10,000 meters above mean sea level. With that much uncertainty in elevation, the temperature estimates are of no value.
Strunk’s point is that the ice melt in Greenland was extremely fast (geologically speaking) and hard to model.
Strunk et al.:
https://www.frontiersin.org/journals/earth-science/articles/10.3389/feart.2018.00129/full
I might add:
The Agassiz record by itself is not a good indicator of temperature changes over all the Greenland, that is why Vinther combined it with Renland. The Arctic has a mind of its own and does not always correlate with global temperature, thus the “warm arctic, cold continents” weather pattern (see Cohen, J., J. Jones, J. C. Furtado, and E. Tziperman (2013).).
The GrIS evolution inferences rely on the Huy3 model, which later variants (e.g., Huy3 Ngr) refined for northwest Greenland. A 2025 study modeling Holocene elevations finds mismatches with Agassiz-derived temperatures, suggesting alternative transfer functions (linear vs. quadratic) for δ¹⁸O could yield different anomalies (e.g., 7°C vs. lower peaks).
It’s a good study, but has flaws and unanswered criticisms.
Thanks for the article, Andy. It’s a nice change from all of the tabloid stuff on this site. It would be nice if WUWT would have a separate section for high-quality articles like yours so that your efforts are visible for more than one or two days.
Proxies should never be combined as climate is local. Using the observation that climate largely repeats after 3560 years, we can compare a proxy with itself and observe that rapid temperature changes are the norm, and that records may be less noisy than they appear. In fact, I’ve moved away from using temperature proxies because they’re often over-filtered; a smooth results appear to have less uncertainty. As I’m more interested in climate periodicity, the high-resolution δ¹⁸O records work better for me.
In the first graphic I compare three δ¹⁸O datasets to each other, and to themself shifted by 3560 years. NGRIP, GISP2 and Vinthor (Agassiz). The most prominent feature, the 8.2ka event is misaligned in the Vinther data. If these three proxies were combined, we would not obtain a better result. I’m not a fan of the Vinther’s et al.. 2009 reconstruction.
Here’s the Kaden et al. 2024 data. Note there’s less filtering on the GISP2 data then for NGRIP and NEEM, yet there’s enough that the rapid 4.7ka event in the δ¹⁸O data has been dispersed over time.
Interesting idea.
Thanks, Andy. A most interesting and informative analysis.
For those interested in the problem with selecting proxies based on their correlation with modern temperatures, see my post
https://wattsupwiththat.com/2024/06/29/mining-for-hockeysticks/
Best to all,
w.
Great post Willis! McIntyre and Wegman et al. each did something similar. This crazy idea of selecting proxies according to their correlation to a short period at the end of their records persists even though it has been shown to introduce a hockey stick bias time and time again. Crazy, but some people just do not learn. Anyway thanks for explaining it so well.
Thanks, Andy. The difference is that McIntyre and Wegman showed it mines hockeysticks out of red noise using principal components, whereas I showed that it mines them using simple averaging.
w.