By Andy May

In 2000, Richard Alley released an ice core temperature reconstruction for Central Greenland using Oxygen isotope ratios. He describes the technique used here. I used this ice core proxy data in a previous post “Climate and Civilization for the past 4,000 years.” Since Alley’s data stops at 1855, I spliced the Greenland HadCRUT surface temperature data on the end to show the Modern Warm Period. In the previous post I erroneously thought Alley’s reconstruction stopped at 1905, not at 1855, so the plot in the earlier post is shifted 50 years. This problem is corrected in this post. I spent too little time on the splice, since my focus was on the effect of climate on civilization and several commenters rightly objected to the error. These included David Middleton, who offered up a better Central Greenland reconstruction by Kobashi, et al. as an alternative. A little research and I was able to find a 4,000 year reconstruction to 1993 by Kobashi et al.

Kobashi uses Argon and Nitrogen isotopic ratios from air bubbles in the ice to estimate paleo-temperatures. He claims that that his method is more accurate than the Oxygen isotope ratios Alley used. Details of Kobashi’s methods are in his PhD thesis, which can be seen here. His unsmoothed results from 2,000 BC to 1993 are shown in Figure 1.

Figure 1

It certainly has a higher resolution than Alley’s dataset. It also has major differences at 700AD and 400BC. Figure 2 is the Alley dataset and the HadCRUT 4.4 surface data corrected for the 50 year error in my previous post. The HadCRUT 4.4 central Greenland temperature anomaly record (in red) is adjusted to the Kobashi record average from 1876-1967 after smoothing both reconstructions. This makes the HadCRUT Greenland temperatures 0.43°C lower than in the previous post. In this plot, the Modern Warm Period is slightly lower than Alley’s Medieval Warm Period.

Figure 2

Figure 3 compares all three datasets. The Kobashi data has been smoothed with a 50 year moving average to make the comparison easier. Alley is unsmoothed and the HadCRUT 4.4 data is smoothed with a 20 year moving average. Both the Medieval and the Minoan warm periods are shifted roughly 50 years between the Alley and Kobashi reconstructions. I take this as the time error (roughly +-50 years) in the datasets. The temperature swings in the Kobashi dataset are more dramatic, but allowing for the time error and resolution differences, generally follow temperature swings in the Alley dataset. Visually, Kobashi’s record has roughly 3x the resolution of the Alley dataset. There are two big differences, the major Kobashi temperature high in 700AD and the major Kobashi temperature low about 550 to 350BC. These differences exceed the apparent error bars of +-50 years and the difference in resolution. Remember both of these datasets were created using the same GISP2 ice cores.

Figure 3

What we can do is go to the historical record to see which record corresponds to history better. Figure 4 is the Figure 1 from my previous post, except I’ve added the Kobashi record and corrected the 50 year error. As before you can click on the figure to get a high resolution PDF.

Figure 4 (click on the image to download a high resolution pdf)

Close inspection of these historical events shows that they fit both reconstructions within acceptable limits. The periods where the timelines differ a lot (400BC and 700AD) have no events. There is also a significant gap at 200BC when the Roman Empire and the Han Dynasty were beginning, but that probably does not disqualify either temperature record.


This is the date where the Kobashi reconstruction hits a peak that is as high as the Minoan Warm Period and over 1.5°C higher than the temperatures we see today. It is also over 3°C higher than the Alley reconstruction at this point. The anomaly lasts 100 years and is precisely opposite of what the Alley reconstruction shows. I’m not sure how to interpret this. This was when Charlemagne unified what is now France and Germany into the Frankish Empire. It was also when the Franks (French) finally stopped the Moslem advance into Europe and held them in what is now Spain. Rashid, sultan of the Abbasid Caliphate, caused the caliphate to reach the height of its power at this time. Great empires are usually built during warm affluent times. I score one point to the Kobashi team.


The Kobashi reconstruction shows this as a very cool time about 200 years long (550 to 350BC). Alley’s reconstruction shows it to be warm. This was the time of Buddha and Confucius. Greece fought off Darius at Marathon and Xerxes fought the “300 Spartans” and sacked Athens. This was also the age of Pericles and when democracy was invented. The Huns begin to invade China, the Romans built their first road. Just at the end of this period Alexander the Great defeated Darius III at Granicus. Pretty important times in history. Generally one would think Alley’s estimate of higher temperatures is correct. Score one for Alley’s team.


Except for these two glaring exceptions, both timelines can be made to fit with major historical events. We do have to assume that the time error is +-50 years and that the temperature error is +-0.2 degrees C or larger and that the Kobashi resolution is roughly 3x the Alley resolution. Visually, the resolution of the Kobashi data (one-half wavelength of the unsmoothed data) is about 50 years. The resolution of the Alley data is no better than 150 years.

After taking these factors into account we are left with two problems, the 700AD difference and the 400BC difference. Both periods are pretty well documented in history and the 700AD difference seems to suggest that Kobashi is correct. But, the 400BC difference suggests the Alley reconstruction is correct. I for one, don’t know what to make of this. I suggest that we remain skeptical of both records and realize that as appealing as they are, they are very rough estimates.  Fortunately, both have the same warm and cool periods and both fit most major historical events reasonably well.


Thanks to everyone who commented on my previous post, especially David Middleton who led me to the Kobashi reconstruction. I’m also in debt to Nick Stokes, Ristvan, Andras Gulacsi and David Middleton who caught the 50 year error and suggested I use actual dates on graph axes rather than years before present which are confusing. Anything for clarity. This is how the best science is done. Rapid and insightful debate and discussion. Blogs are great for science, in a way they take us back to the roots of science. Socrates had his gatherings. Mendel and Darwin worked on their own for the love of science and corresponded with like-minded friends. I particularly like the instant feedback of blogs. I’ve published a lot of peer-reviewed papers where I’d nearly forgotten what I wrote before the peer-reviews came back. No fun in that. Plus, the ego factor and bias inherent in the peer-review process is terrible. Blogs are better.