Guest Post by Renee Hannon

This post compares CO2 ice core measurements from Greenland to those from Antarctica over the last millennium. Paleoclimate studies typically use only Antarctic ice cores to evaluate past CO2 fluctuations. This is because the entire Greenland CO2 datasets were deemed unreliable due to chemical reactions with impurities in the ice and therefore have not been used in studies since the late 1990’s. This post will demonstrate that CO2 data from Greenland ice cores have scientific value and respond to key paleoclimate events such as the Little Ice Age and Medieval Warm Period.

Antarctic Ice Core CO2 Trends
Antarctic ice CO2 data is readily available and has been studied extensively (Bauska, 2015, Ahn, 2012, Siegenthaler, 2005 and Rubino, 2019). Most of the focus of recent studies has been on high snow accumulation sites which tend to have higher resolution and less smoothing of the trapped gas age in ice bubbles due to the firn to ice transition. Gas age width and resolution ranges from 10 years in Law Dome ice cores to 65 years in Dronning Maud Land DML. Figure 1 shows CO2 data from Antarctic high-resolution ice cores over the past millennium.

Ahn et al, 2012, compiled CO2 records from the West Antarctic Ice Sheet (WAIS) and compared them to other key datasets such as Dronning Maud Land (DML), and Law Dome. Their study recognizes and discusses elevated CO2 during the Medieval Warm Period (MWP) at 1000 AD, decrease of CO2 around 1600 AD during the Little Ice Age (LIA) and the subsequent rapid increase beginning around 1850 AD.

Figure 1: Antarctic ice core CO2 data during the past 1000 years. Actual data points are plotted with 20-yr trend lines fitted to these data. Law Dome trend is the 20-year spline provided by Rubino, 2019. Siple and Adelie Land D47 57 is a 100-year trendline from Barnola, 1995. Gas width due to firn-ice transition and sample spacing resolution in years (yr) are noted in text box. Data references shown on plots.

More recently, Rubino 2019, presented revised ice core gas records for Law Dome. His robust study evaluates multiple gases such as CO2, CH4, N2O as well as carbon isotopes. He has a good discussion on the LIA where the ice record shows CO2 decreasing as delta 13C increases which favors reduced soil or terrestrial respiration in response to cooling. Rubino also discusses the CO2 decrease of 10 ppm in the Law Dome record around 1610 AD which is not present in any other records. This rapid decrease demonstrates the higher resolution of Law Dome which is the highest Antarctic snow accumulation site.

Figure 2 compares the Antarctic CO2 trends. Surprisingly, there is quite a bit of difference ranging from 0-11 ppm. Ahn, 2012 noted that WAIS is generally 2-4 ppm higher than Law Dome CO2 data. WAIS CO2 is also systematically higher than DML CO2 by up to 6 ppm, with an average of 4 ppm higher. This shift occurs with both CO2 and its carbon isotopes but not with CH4, methane. The reason for the shift is not well understood.

Figure 2: Antarctic ice core CO2 trends during the past 1000 years compiled from Figure 1. Data references in figure 1.

The high-resolution Law Dome data can identify CO2 rises and dips of less than 30 years. WAIS is up to 10 ppm higher during the Law Dome dip at 1610 AD; due to lower resolution as noted by Rubino, 2019. Law Dome also shows modest CO2 rises during the MWP, but not as high as WAIS. Interestingly, Law Dome is 2-4 ppm higher than DML pre-1600 AD except for a few dips.

DML CO2 has the lowest snow accumulation and is the lowest resolution Antarctic dataset shown. The resolution is about 65 years with a more muted representation of atmospheric CO2 history, explaining the systematic lower CO2 readings. DML does show key CO2 events such as a decrease during the LIA, subtle increase during the MWP, and a rapid increase beginning around 1850 AD. DML and Law Dome CO2 trends converge during the LIA and recent rapid rise.

The Siple D47 D57 data shown as the dashed line in Figure 2 was not utilized in either Ahn or Rubino’s studies due to uncertainty in age dating and/or imprecise experimental methods. Clearly it is the odd dataset, particularly pre-1400 AD. There is an 11-ppm difference in CO2 from Law Dome at 1200 AD. The oddities of the D47 D57 measurements are important to note because this is the Antarctic dataset used to compare and discredit Greenland CO2 data (Barnola, 1995).

Greenland Ice Core CO2 Trend
CO2 measurements from Greenland ice cores are believed to be unreliable due to in situ production of CO2 by carbonate-acid reactions and oxidation of organic compounds (Anklin 1995, Barnola 1995, and Tschumi 2000). This premise was put forward because Greenland CO2 data differed from Antarctic CO2 ice core data. CO2 concentrations in Greenland ranged up to 20 ppm higher than Antarctic CO2 although the records are in good agreement for about the last 300 years. Greenland CO2 had more variability with standard deviations of 6-10 ppm compared to 2-3 ppm in Antarctic ice cores. Also, of note, Greenland CO2 concentrations from ice cores (Summit, GISP2, GRIP, Dye3) agree well with each other and all show similar disagreements from Antarctic. See my previous post for a more thorough discussion of Greenland CO2 data here.

Let’s take a closer look at Greenland ice core CO2 data during the past 1000 years. Barnola, 1995, analyzed ice core samples from Greenland Summit at two different laboratories, Grenoble and University of Bern. Summit has high snow accumulation rates, and the smoothing is only about 15 years. Digital data is not available; however, tables of the data are included in their publication. Sample spacing is erratic and ranges from 5-80 years with an average of 30 years. Carbon isotope data from Greenland ice cores are not publicly available.

Barnola, 1995, states the agreement between the Greenland Summit CO2 measurements from the two laboratories is very good with the mean difference being about 2 ppm. It is common to see 3 ppm discrepancies in CO2 between different laboratories according to Rubino, 2019. Due to the good agreement, Greenland CO2 samples at the same depth were averaged between laboratories. The data was then resampled and smoothed over 60 years. The results are plotted in Figure 3.

Figure 3: Graphs are Grenoble and University of Bern lab CO2 measurements from the Summit Eurocore in Greenland overlain on a low pass filter of 60 years, solid green line. Vertical error bars are 5 ppm deviation. Triangles represent the 2 data points outside error bars in bottom graph.

Greenland ice CO2 decreases to 280 ppm during the LIA and shows a rapid CO2 increase starting about 1850. Greenland CO2 data shows two earlier increases: a better defined Medieval Warm Period and another distinct rise around 1550 AD. CO2 data reaches nearly 300 ppm on individual data points during the MWP.

CO2 Inter-Hemispheric Differences
One of the reasons Greenland CO2 measurements in ice were questioned was due to a large inter-hemispheric difference (IHD) of 20 ppm when compared to Antarctic data. The modern inter-hemispheric gradient of atmospheric CO2 concentrations after being de-seasonalized range from 1-6 ppm measured during the short window of the past 45 years.

The CO2 trends between Greenland and Antarctic data are shown in the top graph in Figure 4. Law Dome is not plotted due to its higher resolution as WAIS and Siple D47 57 mostly bracket the range of Antarctic CO2 data.

Greenland CO2 was originally compared to Antarctic Siple D47 D57 and demonstrated differences up to 20 ppm, which is unreasonable according to Barnola, 1995. Indeed, the bottom graph in Figure 4 shows that the difference between Siple D47 57 and Greenland (gold line) is up to 20 ppm briefly around 1200 AD. For the most part, the IHD is less than 5 ppm from 1300 AD to the LIA, with an exception at 1550 AD. Greenland and Antarctic differences are practically zero during the LIA to present day. The IHD tends to become higher during the MWP and around 1550 AD. As previously mentioned, the Siple D47 D57 ice core is an odd outlier Antarctic CO2 dataset.

Figure 4: Top graph shows Greenland and Antarctic ice core CO2 trends over the past 1000 years. Bottom graph shows interhemispheric difference (IHD) between Greenland and Siple D47 57 in gold and the Greenland and WAIS difference in blue. The light gray shaded band is the acceptable modern IHD range. Error bars represent 2 standard deviations. Modern observatory Barrow (BRW) and South Pole (SPO) IHD data are shown by the gray line.

If the Greenland CO2 trend is compared to Antarctic WAIS ice core, then the interhemispheric difference is always less than 10 ppm shown in blue in Figure 4. And it’s less than 5 ppm during 70% of the past 1000 years like the present atmospheric CO2 differences (the heavy gray line). It should be emphasized that Antarctic datasets have up to 10 ppm difference between them alone. Thus, comparing Antarctic and Greenland datasets and discovering a difference of 5-10 ppm appears to be within the range of reasonable values considering their geographic distance.

The highest polar inter-hemispheric difference and the highest difference between the Antarctic CO2 datasets both occur during the MWP. The D47 D57 shows the greatest divergence from the Antarctic datasets but it is not used in recent publications such as Rubino, 2019 or Ahn, 2012. Unfortunately, the Greenland CO2 data was originally compared to this outlier Antarctic D47 D57 dataset and deemed unacceptable.

Interestingly, all Antarctic and Greenland CO2 measurements tend to converge during the colder LIA period and overlie almost perfectly during the subsequent rapid increase.

Greenland Ice CO2 Rises Mimic Methane
Greenland and Antarctic CO2 trends plotted alongside methane from ice cores over the past millennium are shown in Figure 5. Methane from various ice cores overlie nicely. Methane shows a distinct separation between the polar regions with an IHD of 24 to 58 ppb (Rubino, 2019). This is lower than the current atmospheric methane difference of 100 ppb between the South Pole and Barrow observatories.

In general, methane shows similar trends over the past millennium as CO2. Methane decreases during the LIA with a subsequent rapid increase. There is a distinct rise in methane around 1550 AD that is not captured well by Antarctic ice CO2, especially the DML data. Interestingly, the Greenland CO2 does show the 1550 AD increase. Greenland CO2 also shows character during the MWP that mimics the methane trends, again not seen in the Antarctic CO2 data.

Figure 5: Top graph are Greenland and Antarctic ice core CO2 trends over the past 1000 years. References in Figure 1. Bottom graph shows methane data from Greenland (green/grays) and Antarctic (red) ice cores. GISP2 and WAIS methane from Mitchell, 2013 and 2011; NEEM methane is from Rhodes 2014.

Antarctic CO2 shows more scatter than methane in the various ice cores over the past millennium. As discussed above, Antarctic WAIS is 2-4 ppm higher than Law Dome and 3-6 ppm higher than DML. Additionally, the lower resolution Antarctic DML is 2-4 ppm lower than Law Dome during the MWP and 1550 event. CO2 from all datasets tend to converge during the LIA cold period and subsequent rapid rise.

Figure 6 illustrates some of these CO2 differences for three key events: MWP, 1550, and LIA. All datasets recognize these events; however, the magnitude between the events varies. The left graph shows that Greenland and WAIS have higher average CO2 concentrations during all events. Greenland and WAIS were normalized on the LIA which required a shift of 3 ppm shown on the right graph. Greenland shows the largest magnitude, or CO2 amplitude variation, between the cooler LIA and MWP of over 11 ppm. As expected, DML shows the lowest difference of only 2 ppm between the LIA and MWP due to its lower resolution. Law Dome shows a slightly higher difference of 5 ppm compared to WAIS of 4 ppm.

Figure 6: Average CO2 from Greenland and Antarctic trendlines for key events. Left graph is trendline averages and right graph is normalized on the LIA. Age ranges used for MWP from 1000-1200 AD, 1550 event from 1500-1660 AD, and LIA from 1700-1800 AD.

The reasons for the CO2 scatter and different underlying trends are not well understood and chemical reactions within the core are frequently cited. The scatter in both Greenland and Antarctic tends to occur with elevated CO2 during warmer times.

Another potential explanation is the modification of CO2 during the firn to ice transition. Ahn, 2012, states that WAIS CO2 probably experience additional smoothing processes not captured by firn air models. Their enhanced firn air model still underestimates WAIS CO2 smoothing by 37 percent (19 years versus 30 years). Temporal resolution is certainly a factor in the systematically reduced CO2 measurements in the DML ice core.

Scatter in CO2 may be also be attributed to inter-core variability. Rubino notes that it is not uncommon to see inter-core variability of 3-4 ppm in Antarctic core data. Greenland’s present day atmospheric data shows higher standard deviations from 4-5 ppm with 15-20 ppm seasonal swings than Antarctic atmospheric data. It should not be a surprise that inter-core CO2 variability is higher in Greenland ice cores than Antarctic cores.

Scientists have classified Greenland ice core CO2 measurements as contaminated and mostly ignore these data. These data are a high-resolution polar endmember that can provide additional information to complement Antarctic CO2 datasets. Higher CO2 events are expressed better in Greenland than Antarctic cores. For example, Greenland CO2 data captures CO2 increases almost up to 300 ppm in individual data points during the MWP that correlate well with methane rises. Greenland CO2 data show increases during 1550 AD like methane rises. Perhaps, Greenland data is suggesting that CO2 increases during past warm periods are larger than documented by the muted Antarctic CO2 data. Greenland CO2 data is trying to tell scientists the Arctic side of the paleoclimate story. But many scientists have chosen not to listen.

Errors using inadequate data are much less than those using no data at all” – By Charles Babbage, Inventor and Mathematician

Acknowledgements: Special thanks to Donald Ince and Andy May for reviewing and editing this article.

Download bibliography here.


Andy May, now retired, was a petrophysicist for 42 years. He has worked on oil, gas and CO2 fields in the USA, Argentina, Brazil, Indonesia, Thailand, China, UK North Sea, Canada, Mexico, Venezuela and Russia. He specializes in fractured reservoirs, wireline and core image interpretation and capillary pressure analysis, besides conventional log analysis. He is proficient in Terrastation, Geolog and Powerlog software. His full resume can be found on linkedin or here: AndyMay