Contrarian Manipulation of Paleoclimate Data

 

Now that 2023 is in the books, it's pretty clear that all global temperature datasets will show 2023 to be the warmest year on record by a significant margin. And since 21st century has exceeded the warmest temperatures of the Holocene (the current interglacial), it's also fair to say that 2023 has likely been the warmest year at least since the last interglacial ended 125K years ago. I'm going to wait until all the major datasets are updated to 2023 before saying more about 2023. But already the dishonesty from contrarians is coming out to spin 2023 into something other than what it actually was.

Patrick Moore's Dishonesty

Patrick Moore posted a tweet with the above graph claiming that what I summarized about 2023 above is an "outright lie." His words:

The claim is being made that “2023 was the hottest year in 125,000 years”. This is an outright lie. The Holocene Climatic Optimum from about 10,000-5,000 years ago, when the Sahara was green, was warmer than this Modern Warm Period.

But if the graph he shared was intended to demonstrate that 2023 isn't the warmest of the Holocene, it failed. In fact, the graph is a convoluted mess. The CH4 time series is attributed to Kobashi et al 2007, but it looks like it actually comes from either from Grachev et al 2007[1] or Monnin et al 2004[2]. I suspect both the CO2 and CH4 time series come from the latter. The "model temperatures" are strangely put on a different temperature scale from the proxy temperatures, but that time series actually does come from Liu et al 2014, which was a very important paper on the "Holocene Temperature Conundrum." The paper discusses the discrepancy between the models (which generally show slowly increasing temperatures throughout the Holocene with increasing CO2) and Marcott's reconstruction of global temperatures in the Holocene published 2013.[6]

Liu et al 2014

The conundrum is likely caused by seasonal biases affecting the proxy reconstructions (a suggestion made in Marcott's paper), and this has been understood for some time and discussed in the literature. Recently, Samantha Bova developed a technique to correct for these biases in reconstructions[4], and newer reconstructions like the reconstruction by Osman et al 2020[5] shows the Holocene temperatures to be closer to the those expected by models and CO2 proxies.

Patrick Moore doesn't tell you any of this, though. Instead he shows a "proxy temperature" time series that he attributed to Marcott et al 2013. The problem, though, is that this time series doesn't come from Marcott. It's actually a manipulated version of Marcott's proxy data created by Javier Vinós. Vinós took Marcott's proxies and manipulated them to make the HTM warmer. In Marcott's reconstruction (see below) warmest and coolest portions of the preindustrial Holocene differ by about half the amount of Vinós' version. And of course, all the time series in Moore's graph end in 1950. The instrumental record is missing, so his graph is missing the last 73 years of global warming. It's amazing to me that he thinks he can show that it's a "lie" that 2023 is the warmest of the last 125K years while displaying a graph that omits the last 73 years of global warming.

So let's test Moore's claim that it's an "outright lie" that 2023 was the warmest of the last 125K years. Below I plotted the correct Marcott 2013 set to the 1850-1900 mean along with HadCRUT5 with 20-year smoothing to match the resolution of Marcott's reconstruction. I also included the 95% CIs for Marcott's reconstruction. There's almost certainly more variability in GMST in the Holocene than shown in Marcott's reconstruction (discussed further here). But clearly the 2014-2023 mean is about 0.25°C warmer than the upper CI for the warmest years of the Holocene. There is no evidence of any 20-year period of the Holocene matching 2014-2023, and since the Holocene interglacial was warmer than the glacial period that preceded it, the claim that 2023 was likely the warmest year of the last 125K years stands.

Marcott 2013

Javier Vinós' Dishonesty

But there's more to this story. I think it's important to examine how Javier Vinós tampered with the proxy data in Marcott 2013 to make the HTM too warm in Moore's graph. Here's another version (below) credited to Javier Vinós' self-published book that allows you to see the extent to which Vinós tampered with Marcott's proxy data. Here Vinós used the differencing method to average the proxy data, a method that he credits following Tamino to account for proxy drop off at the end of the time series (essentially the first half of the 20th century). Vinós' version is the black curve (b) below, plotted as a Z-score. The graph also shows Marcott's Standard 5x5 time series as the red curve (a), which he says correct for proxy drop off and produces an artificially enhanced spike. That's true (and discussed by Marcott), but Vinós didn't mention that Marcott's RegEM does account for proxy drop off. Vinós could have used this reconstruction if he chose. The discussion in Tamino's post (which I discuss here) had to do with whether the difference between the Standard 5x5 and RegEM during the 20th century was robust. Tamino argued that it would be robust if Marcott used the differencing method for the Standard 5x5. But Vinós plotted his version of Marcott's data as a Z-score (black) and Marcott's Standard 5x5 as temperature (red) and then scaled the two curves differently to make the HTM appear warmer in Vinós' curve than in Marcott's.

Already we can see the dishonesty in presentation in Vinós' graph, but he also takes this a step further. In other places, as on Judith Curry's blog, Vinós eliminated the Z-Score scale and replaced it with a temperature scale that inflates the warmth of the HTM. I'm not making this up. He admitted to doing precisely that. He says that the "temperature anomaly was rescaled to match biological, glaciological, and marine sedimentary evidence, indicating the Holocene Climate Optimum was about 1.2°C warmer than LIA." This is nonsense. The proxies used by Marcott already included included "biological, glaciological and marine sedimentary evidence," and that evidence showed that the HTM was only about 0.7°C warmer than the LIA. It's patently obvious that Vinós did not rescale his graph to match this evidence but to deviate from it.

Map of "biological, glaciological, and marine sedimentary evidence"
used in Marcott 2013, from Figure S1

Vinós tries to justify his data tampering by saying that there is "vast literature" on his side. He says, "I have also rescaled the temperature changes to make them congruent with the vast literature and consilience of evidence from different fields that indicates that the Holocene Climatic Optimum was on average between 1 and 2 °C warmer than the Little Ice Age." So Vinós read somewhere that the HTM should be globally 1°C to 2°C warmer than the LIA, and because the actual temperature data didn't match what he believed it should be, he just changed the data to match his expectations. And where is the "vast literature" showing that the HTM was globally 1°C to 2°C warmer than the LIA ? He predictably doesn't say. Vinós essentially admits to data tampering here. Marcott's data didn't match what he believed it should show, so he changed the data. The Vinós-tampered version doubles the difference between the HTM and the LIA, changing that difference from 0.7°C to 1.5°C.

Compare the a and b curves for the extent of Vinós' data tampering

Conclusion

The graph shared by Moore is also in Vinós' post on Curry's blog. I've seen several versions of this graph in various places, including Andy May's blog and in a manuscript by Nicola Scafetta. There is plenty of dishonesty involved here, and Moore and Vinós are not the only people to blame. But Moore (and other contrarians) want us to think of them as skeptics. Even if Moore wasn't aware of Vinós' dishonesty, at best he wasn't skeptical here. He took a dishonest graph and used it to make an invalid point. And anyone can see that the graph ignored the last 73 years of global warming; it does not even place 2023 on the graph. In point of fact, seasonal biases affecting Marcott et al 2013 probably make the HTM too warm, but even if we use Marcott instead of Osman, temperatures have increased above the warmest temperatures of the Holocene.

References:

[1] Grachev, A. M., E. J. Brook, and J. P. Severinghaus (2007), Abrupt changes in atmospheric methane at the MIS 5b–5a transition, Geophys. Res. Lett., 34, L20703, doi:10.1029/2007GL029799.

[2] Eric Monnin, Eric J Steig, Urs Siegenthaler, Kenji Kawamura, Jakob Schwander, Bernhard Stauffer, Thomas F Stocker, David L Morse, Jean-Marc Barnola, Blandine Bellier, Dominique Raynaud, Hubertus Fischer. Evidence for substantial accumulation rate variability in Antarctica during the Holocene, through synchronization of CO2 in the Taylor Dome, Dome C and DML ice cores. Earth and Planetary Science Letters, Volume 224, Issues 1–2 (2004): 45-54. https://doi.org/10.1016/j.epsl.2004.05.007.

[3] Liu, Z., Zhu, J., Rosenthal, Y., Zhang, X., Otto-Bliesner, B. L., Timmermann, A., … Elison Timm, O. (2014). The Holocene temperature conundrum. Proceedings of the National Academy of Sciences, 111(34), E3501–E3505. doi:10.1073/pnas.1407229111.

[4] Bova, S., Rosenthal, Y., Liu, Z. et al. Seasonal origin of the thermal maxima at the Holocene and the last interglacial. Nature 589, 548–553 (2021).
https://doi.org/10.1038/s41586-020-03155-x

[5] Osman, M.B., Tierney, J.E., Zhu, J. et al. Globally resolved surface temperatures since the Last Glacial Maximum. Nature 599, 239–244 (2021). https://doi.org/10.1038/s41586-021-03984-4

[6] Marcott, Shaun et al. “A Reconstruction of Regional and Global Temperature for the Past 11,300 Years.” Science 339 (2013): 1198-1201. http://shpud.com/Science-2013-Marcott-1198-201.pdf