Global versus Greenland Holocene Temperatures

By Andy May

Last week, I posted a global temperature reconstruction based mostly on Marcott, et al. 2013 proxies. The post can be found here. In the comments on the Wattsupwiththat post there was considerable discussion about the difference between my Northern Hemisphere mid-latitude (30°N to 60°N) and the GISP2 Richard Alley central Greenland temperature reconstruction (see here for the reference and data). See the comments by Dr. Don Easterbrook and Joachim Seifert (weltklima) here and here, as well as their earlier comments.

Richard Alley’s (Richard Alley, 2000) central Greenland reconstruction has become the de facto standard reconstruction and is displayed often in papers and posts. And, truth be told, I’ve often used it. See here for an example. But, it is a central Greenland reconstruction, uncorrected for elevation differences over time, and all of Greenland is north of 60°N. A better comparison is with my Arctic reconstruction that goes from 60°N to the North Pole. This comparison is shown in figure 1.

Figure 1

Alley’s reconstruction is based upon trapped air in ice cores taken from central Greenland and his proxies are calibrated to air temperatures on land. My Arctic reconstruction is based upon nine proxies, five are marine proxies and 3 are land proxies. Only one of the land proxies is a Greenland ice core and I used a composite of two Greenland area ice cores, Agassiz and Renland, by Vinther, et al. (2009) and not the better-known Alley reconstruction. The Vinther reconstruction and the Alley reconstruction are compared, using actual temperature, in figure 2.

Figure 2

As can be seen in figure 2, the Vinther Agassiz and Renland reconstruction is less erratic and has a more prominent Holocene Climatic Optimum (HCO) than the Alley reconstruction. In addition, the Vinther Medieval Warm Period is older and the Roman and Minoan Warm periods are far less prominent and offset in time. Notice the reconstructions match in the Little Ice Age (LIA) and that the Vinther Holocene Climatic Optimum (HCO) from 8000 BC to 4500 BC is more prominent. The HCO doesn’t really show up in the Alley record. Below we compare our Arctic reconstruction to the Vinther record in Figure 3.

Figure 3

Vinther’s record shows a more prominent HCO than ours, more detail and a deeper LIA. Finally, let’s compare both Vinther and Alley to our Northern Hemisphere mid-latitude reconstruction in figures 4 and 5.

Figure 4

It is interesting that Vinther agrees with the mid-latitude Northern Hemisphere reconstruction in the Neoglacial period (roughly 5700 BP or 4300 BCto the present), but agrees better with the Arctic reconstruction during the HCO. I’m not completely sure why that is.

Figure 5

Comparing figure 4 to figure 5, we can see that Alley has a very flat trend and is more active than Vinther. Vinther is a better match to our Northern Hemisphere mid-latitude reconstruction. Alley’s reconstruction starts to show the HCO and then fizzles at about 1,000 years in to it. Figures 4 and 5 are anomalies from the mean temperature from 9000 BP to 500 BP, however, which distorts the picture a bit given the two reconstructions differ on the temperatures of the HCO and the LIA. I refer you to figure 2, where we compare Vinther to Alley in actual temperature and not in an anomaly form. Here the two reconstructions agree on the temperature of LIA, but the Alley reconstruction does not see the HCO. We see that the key difference between the two is the degree of warming during the HCO.

Why are Alley and Vinther different?

The short answer is that Vinther, et al. (2009) corrected their ice core records, including GISP II and GRIP, for elevation differences and Alley did not. In Vinther’s words:

“The previous interpretation of evidence from stable isotopes (δ18O) in water from GIS [Greenland Ice Sheet] ice cores was that Holocene climate variability on the GIS differed spatially and that a consistent Holocene climate optimum—the unusually warm period from about 9,000 to 6,000 years ago found in many northern latitude palaeoclimate records—did not exist. Here we extract both the Greenland Holocene temperature history and the evolution of GIS surface elevation at four GIS locations. We achieve this by comparing δ18O from GIS ice cores with δ18O from ice cores from small marginal icecaps [Agassiz and Renland]. Contrary to the earlier interpretation of δ18O evidence from ice cores, our new temperature history reveals a pronounced Holocene climatic optimum in Greenland coinciding with maximum thinning near the GIS margins. Our δ18O -based results are corroborated by the air content of ice cores, a proxy for surface elevation.”

In figure 6 we see a summary of the Vinther, et al. (2009) data, it is their figure 1.

Figure 6 (Source: Vinther, et al. 2009)

The six cores are well distributed across Greenland, with Agassiz on Ellesmere Island very close to Greenland. Agassiz and Renland are both coastal cores and have similar profiles. It is possible to reconstruct the elevation histories for these two locations with confidence, so they are used to develop corrections for the remaining 4 ice cores. All six core records shown were included in the Vinther, et al. (2009) reconstruction after adjustment for elevation and ice thickness changes, but the Agassiz and Renland cores are the key cores. The corrections to these cores are shown in 6D. The δ18O profiles for these cores, after the uplift (or elevation) correction has been applied, is shown in 6c. Considering that Agassiz and Renland are on opposite sides of the GIS and 1,500 km apart, the agreement between the two corrected records is astounding, as Vinther, et al. (2009) described it in their paper.

Alley’s reconstruction focused on the GRIP and GISP II cores, these two cores are 30 km apart in central Greenland, they are combined into one point called GRIP in figure 6.

Below is a better location map for the Greenland ice cores, shown as figure 7.

Figure 7 (Source CDIAC)

Temperature determination in these ice cores is done with a function of δ18O and it has been shown by Johnsen and White (1989) that the average δ18O level over and around the Greenland Ice Sheet (GIS) is almost completely described by altitude (-0.6‰/100m) and latitude (-0.54‰/degree N). The altitude effect is due to the moist-adiabatic cooling of an air mass rising above the GIS. As it cools, precipitation and fractionation take place. There are more details on this in the Vinther, et al., 2009 supplementary materials. Thus, there is a sound basis for building a good δ18O temperature record if the altitude of the ice surface is known throughout the Holocene. Elevation differences must be taken into account. As Vinther, et al. (2009) write:

“… the differences in the long-term δ18O trends seem to be related to changing GIS elevation …”

The Holocene Climatic Optimum was a warm period and it caused melting of the GIS. Thinning at the Camp Century and DYE-3 sites started very soon after the HCO began over 9,000 years ago. The thinning progressed from there to the GISP II/GRIP location in a few thousand years, certainly by 6,000 BP. This affected the GISP δ18O temperature record and all but eliminated the HCO response that we see in other Northern Hemisphere records. The elevation corrections applied to the four sites, including GRIP, NGRIP and GISP II are shown in figure 8, from Vinther, et al. (2009).

Figure 8

The Camp Century and DYE-3 locations are on the coast and they are affected most. In the interior GRIP and NGRIP locations (remember GISP II is next to GRIP, see figure 7) the effect is less, but still significant.


If we accept the work that Vinther, et al. (2009) have done as being correct, and I see no problems with it, then the Aggasiz-Renland δ18O records, after correction for elevation records are correct. These records are 1,500 km apart and on opposite sides of the GIS, thus the temperature record of Greenland for this period must be fairly uniform for this period of time. Because of the geological conditions at the Aggasiz and Renland sites, their elevation histories can be reconstruction with some confidence as explained in Vinther, et al.’s paper and supplementary materials. Given that we also know the controls on the average δ18O with confidence, then we can provide a reliable temperature record for these sites. This is the record used in my Arctic reconstruction and the other 8 records used in the reconstruction agree fairly well.

Vinther, et al.’s reconstruction also agrees well with my Northern Hemisphere reconstruction from 4,000 BC to the present. It reaches a lower temperature extreme in the HCO, but matches the HCO of my Arctic reconstruction. Generally, I prefer the Vinther et al. reconstruction to Alley’s earlier GISP II reconstruction for the purpose of detecting the major climatic events of the Holocene and estimating the difference between HCO temperatures and LIA temperatures.

However, for locating climate events in time and whether the event is a warming event or a cooling event, using a single ice core proxy, that is well dated is fine. And the dating error in ice cores is very low, less than 1% (Alley, 2000). It is just that the magnitude of the temperature swings are probably incorrect in the GISP II and GRIP cores due to elevation changes as Vinther, et al. have shown. These changes (or errors in temperature) are the most severe in the HCO. This problem affects the magnitude of the estimated temperature but not the timing of the events.

Using multiple proxies, as I have, helps measure a more accurate and robust temperature anomaly for a region or the whole globe, but adversely affects the timing of events due to averaging multiple proxies with possibly inaccurate dates. Dating errors of 100 to 150 years are probably common and when averaging records with this sort of error, there will be loss of short term amplitude and problems estimating the timing of major events. This always needs to be considered in this sort of work. Amplitude reduction or excessive smoothing of the temperature reconstruction can be minimized by using fewer proxies, higher resolution proxies (shorter sample intervals), minimizing the proxy drop out at both ends of the reconstruction by avoiding short term proxies, and selecting proxies that are not overly affected by local events or local geology. Careful proxy selection is critical for a robust record, for more details on proxies to be avoided and proxies to include see my posts on the reconstructions I made. The final post, which will lead you to the others is here.

So, what is the purpose? Do you want to know, as accurately as possible, when a Northern Hemisphere warming event or cooling event occurred? Then using GISP II or GRIP will work best. Do you want to estimate the average temperature change during the event? Then I would recommend my reconstructions, but realize that the estimate may be conservative and the date of the event may be incorrect by 100 to 150 years. Our knowledge and data about Holocene temperatures are limited, but by using what we have wisely we can begin to get our arms around it.

Published by Andy May

Petrophysicist, details available here:

Leave a Reply

%d bloggers like this: