About the reliability of ice cores…
Guest Post by Ferdinand Engelbeen
There have been hundreds of reactions to part 1 about the mass balance (http://wattsupwiththat.com/2010/08/05/why-the-co2-increase-is-man-made-part-1 ). Many respondents still are not convinced that the mass balance is a firm proof that the observed increase of CO2 in the atmosphere is human made. But there are more indications. Ultimately, any alternative explanation must fit all the observations. If the alternative hypothesis fails even only one of the observations, then the alternative is rejected. But before we start to look at more observations which support an anthropogenic cause, we need to address several misconceptions which fly around on the Internet, mainly on skeptic blogs… This part has a detailed look at the reliability of ice cores, which are quite important for our knowledge of the pre-industrial CO2 levels, but have been subject to a lot of critique.
Note that the ice cores only show CO2 levels back to about 800,000 years, but measurements may in the future be extended to over one million years. What is found in the ice cores is only relevant for the most recent period of our history and not for more distant geological time periods.
About the reliability of ice cores:
Some have objections to the ice core measurements, as these are regarded as the main reason for the “equilibrium” assumption of ancient CO2 levels. The only real problem in this case is the smoothing of CO2 levels. That depends on the snow accumulation rate, as it takes a lot of time to close all air bubbles in between the snow flakes. That happens at a certain depth where the pressure is high enough to transform the snow, then firn (densified snow still with open pores) into ice. The averaging happens partly because at first the firn pores are large enough to let the air in the pores and in the atmosphere exchange with each other, partly because some bubbles close early, others at a lower depth (thus contain air which is different in composition, “age”, than other already closed bubbles). The depth where this happens depends on the pressure from the layers above and the temperature of the ice. The time needed for full closure of all bubbles largely depends on the accumulation rate of snow at the place where the ice core is taken (or upstream if coring at a slope).
That makes that the average smoothing of CO2 levels is about 8 years (Law Dome 2 out of 3 ice cores, 1.2 m ice equivalent/year accumulation), some 21 years (the third Law Dome ice core, 0.6 m ice equivalent, see http://www.agu.org/pubs/crossref/1996/95JD03410.shtml unfortunately behind a pay wall…), some 570 years (Dome C, a few mm/year, see http://www.nature.com/nature/journal/v453/n7193/full/nature06949.html ) and everything in between. The Law Dome closing period of the bubbles was measured, while for Dome C one needed models to estimate the time resolution in the far past.
Thus the smaller the snowfall at a certain place, the longer it takes for the bubbles to fully close and the longer averaging one has. At the other side, the smaller the accumulation rate, the further we can look back in the past, as for the same depth of ice, there are many more years of snowfall.
The fact that the pores still are open over a long period, also means that there are differences in the age of the ice and the age of the enclosed gas. The age of the ice can be counted, as it simply is the result of ice formation from yearly snow accumulation where winter/summer snow density differences gives clearly distinguishable layers if there is sufficient accumulation. If, as depth increases, the pressure and/or flow result in layers that are near invisible, one may use several other methods like electro conduction or X-rays (see http://iopscience.iop.org/1742-6596/41/1/034/pdf/jpconf6_41_034.pdf ) to distinguish the layers/age.
Determining the gas age is not as easy. Over the years of accumulation of the snow/firn, the pressure builds up and the firn becomes more dense with decreasing pore diameter. That reduces the exchange of air in the pores with the air in the atmosphere, until the pores are too small to make any further exchange possible. If there has been considerable accumulation, as in the two fast Law Dome cores, at the depth of the first closing (about 72 meters) the ice is already 40 years old (40 layers), but the air has the average CO2 levels of less than 10 years ago, which makes the average gas age (including the average time for fully closing of all bubbles) about 30 years younger than the ice at the same depth. For the top layers, we have the advantage of direct measurements in the atmosphere for overlapping periods, which makes a comparison possible.
For cores with far less accumulation, the analysis is more problematic, as the difference increases with the reciprocal of the accumulation rate. During ice ages, there was less precipitation, thus increasing the ice age – gas age difference. The ice-gas age difference for the Vostok ice core is over 3,000 years. Be aware that the ice-gas age difference has nothing to do with the resolution of the CO2 levels, as these are in the bubbles themselves, but it makes a chronology of what happens between temperature (measured as dD and d18O proxy in the ice, see further) and CO2 levels (measured in the bubbles) more difficult to establish. But here also different techniques are used: diffusion speed is a matter of pore diameter, directly related to firn/ice density and densification speed is directly related to accumulation speed. This can be used to model the exchanges between air in the pores and the atmosphere.
The calculations to establish the gas age did fit quite well for the Law Dome ice cores, where besides ice age, the average gas age was established by measuring CO2 levels top down in the firn. That showed that the gas age at closing depth was less than 10 years old on average, but more importantly, the CO2 levels in the already fully closed bubbles and the still open pores were the same. For the low accumulation ice cores like Vostok, there is more discussion about the ice-gas age difference and different time scales were established…
The accuracy of the measurements in the three Law Dome ice cores for the same gas age is about 1.2 ppmv (1 sigma). Later works compared different ice cores for CO2 levels at the same average gas age. These show differences of only 5 ppmv, despite huge differences in average temperature (coastal -20°C, inland -40°C), salt inclusions (coastal), accumulation rate and resolution. There are a lot of overlapping periods between the ice cores, the resolution decreases with increasing length of period (from 150 years – for 2 of 3 Law Dome ice cores – to 800,000 years – Dome C), but even so, the measurements (done by different labs of different organizations) show a remarkable correspondence for the same average gas age. This is a nice indication that the CO2 levels of the ice cores indeed represent the ancient levels.
Data over the past 10,000 years of average gas age in ice cores from:
As result, for the past 150 years (Law Dome) we have accurate data with a reasonable resolution. The cores average the CO2 levels over 8 years, so any peak of 20 ppmv during one year or 2 ppmv difference sustained over 10 years would be observable. For older periods, the resolution is less and the averaging applies to the full period of resolution (about 570 years for Dome C).
The visual correlation between temperature and CO2 levels in ice cores is well known to everybody, as that was used by Al Gore and many others, although he forgot to tell his audience that the CO2 levels lagged by some 800 years during a deglaciation and many thousands of years at the onset of new glaciations:
Data from the Vostok ice core via:
The temperature is derived from dD and d18O proxies in the ice. dD means the change in the deuterium/hydrogen ratio measured in the water molecules of the ice and d18O is the change in 18O/16O ratio of the water molecules in the ice. Both heavier isotopes of hydrogen resp. oxygen increase in ratio to the lighter ones, when the ocean temperature, from where the precipitation originates, increases. Thus the change in ratio is an indication of the ocean temperature changes. For coastal ice cores, that indicates the temperature changes in the nearby Southern Ocean, while the deep inland cores receive their precipitation from the more widespread SH oceans, thus representing the temperature changes of about the whole SH. The NH ocean temperature changes are more or less represented in the Greenland ice cores, which show similar changes (over the last about 120,000 years), but with some differences in timing and more detailed extreme events (like the Younger Dryas).
There is a remarkable near-linear ratio between ice core CO2 and the temperature proxy record in the same core over 420,000 years of Vostok. Work is under way to confirm this ratio in the 800,000 years of Dome C (for the overlapping period, the CO2 levels are already confirmed similar): about 8 ppmv/°C:
Data of the Vostok ice core from NOAA, temperature proxy indication shows zero at current temperature. From:
The spread in temperature/ CO2 data, mainly at the high side, is from the long lag of CO2 levels which remain high for thousands of years at the end of a warm period, while the temperature is dropping back to a minimum. The 8 ppmv/°C is not absolutely right, because temperature at best represents a hemispheric ocean temperature, but not far off, as the pCO2 in seawater dependency of temperature shows about 16 ppmv/°C. But besides pCO2 of seawater, other land and (deep) ocean items also play a role.
This all is an indication that temperature is not the cause of the sharp increase of CO2 in the last 150 years, as that wouldn’t give more than 8 ppmv (or 16 ppmv based on ocean solubility) increase with a maximum 1°C temperature increase since the depth of the LIA, while the current increase is over 100 ppmv.
Be aware that, besides some fractionation of the smallest atoms/molecules (not of CO2), and a small fractionation of isotopes, the bubbles still reflect the ancient atmosphere as it was. Ice core CO2 thus is not a proxy but a direct measurement, be it smoothed, of what actually happened in the (far) past.
The objections of Jaworowski:
What about the objections of Jaworowski against the reliability of ice cores (http://www.warwickhughes.com/icecore/ )?
Jaworowski assumes that CO2 “leaks” via cracks in the ice, caused by the drilling and pressure release of the deep core ice. But how can they measure 180-300 ppmv levels of CO2, when the outside world is at 380 ppmv? If cracks (and drilling fluid) are found in the ice, that would show levels which were too high, compared to other neighbouring layers, never too low.
The formation of clathrates (solid forms of O2, N2 and CO2 with water at very cold temperatures and high pressure) depletes CO2 levels, according to Jaworowski. This is well known in the ice core world. Therefore they allow the ice cores to relax up to a year after drilling. Moreover: O2 and N2 clathrates would decompose first, thus escaping as first via microcracks (as Jaworowski alleges). This would lead to too high CO2 levels, not too low.
Jaworowski accuses Neftel of “arbitrary” shifting the Siple data with 83 years to match the ice core CO2 with the Mauna Loa data. But the page from Neftel’s report ( http://www.biokurs.de/treibhaus/180CO2/neftel82-85.pdf ) contains two columns in the table: the counted ice age and the calculated gas age, the latter based on porosity measurements of the firn. Jaworoski used the age of the ice, not of the air bubbles, to base his accusation on, which is quite remarkable for a specialist in these matters. CO2 is in the air, not in the ice and the average age of the gas is (much) younger than the ice where it is enclosed. Neftel even made specific remarks about the gas age, which was compared to the South Pole atmospheric data, to confirm the average age of the gas bubbles at depth:
If the 328 p.p.m. measured at a depth of 68.5 m.b.s. [note: meters below surface] is matched with the atmospheric South Pole record, the mean gas age is 10 yr, corresponding to a difference between mean gas age and ice age of 82 yr, which lies in the above estimated range. The difference is used in calculating the mean gas age for all depths.
That the CO2 concentration measured on the subsequent samples from 72.5 and 76.5 m.b.s. corresponds with the atmospheric South Pole record justifies this age determination…
This clearly indicates that Neftel based his gas age estimate on firm grounds and there is nothing arbitrary in “shifting” the data, as there was no shifting at all. Thus for the Siple ice core, the ice age – gas age difference is about 82 years (Neftel estimated 80-85 years) for an average gas age resolution of about 22 years in this case.
Many of the objections of Jaworowski were answered by Etheridge (already in 1996) by drilling three cores at Law Dome, with three drilling methods (wet and dry), using different materials for sampling, avoiding cracks and clathrates, allowing a lot of relaxation time and measuring the CO2 levels top down in firn and ice. No difference was found in CO2 levels between firn and ice at closing depth and there is an overlap of some 20 years of the ice core CO2 data with the South Pole data:
Figure from Etheridge e.a.: http://www.agu.org/pubs/crossref/1996/95JD03410.shtml
See more comment and further links about Jaworowski at:
The “corrections” of J.J. Drake:
JJ Drake (http://homepage.ntlworld.com/jdrake/Questioning_Climate/userfiles/Ice-core_corrections_report_1.pdf ) claimed to have established that the CO2 levels needed a correction for the ice-gas age difference. The result of the “correction” is that the CO2 levels are much higher with little variation and the very good correlation with temperature vanished. This conflicts already with our knowledge of the influence of temperature on CO2 levels in current times…
Even so, the “correction” might be all right, but the reason he provided has no bearing in any physical relationship. He makes the basic mistake of conflating a good correlation with a causation: The error is of the kind:
A causes B and shows a good correlation.
A causes C and shows a good corelation.
Thus B causes C, because there is a good correlation between the two. But that correlation is completely spurious, as there is not the slightest physical connection between B and C.
The explanation for his observation is quite simple:
Temperature (“A”) causes the ice-gas age lag (“B”), as temperature is directly connected with humidity of the atmosphere, thus influences the amount of snowfall, thus the accumulation rate and as reciprocal the speed of closing the bubbles: higher temperature, higher snowfall, smaller ice-gas difference.
Temperature (“A”) influences CO2 levels (“C”) directly: higher temperature means higher CO2 levels.
Because the previous two results have a high correlation with temperature, that gives that the ice-gas age difference and the CO2 levels also show a high correlation, but there is no physical mechanism that shows any direct or indirect action of ice-gas age difference on CO2 levels or vice versa. It is a completely spurious correlation, without any causation involved, but both share the same cause. Any “correction” of CO2 levels found in ice cores based on the correlation with ice-gas age difference is meaningless.
Migration of CO2 in ice cores
Ice shows a thin layer of unstructured (liquid waterlike) water molecules near the surface of the air bubbles. Some CO2 may dissolve in this layer, but that is not a problem at measurement time, as measurements are made at low temperature under vacuum, effectively removing all CO2 from the opened bubbles in the crushed ice, while removing any water vapor as ice over an extra cold trap. Water in-between the ice crystals is very unlikely, as there is still the direct influence of ordered structural ice from both sides.
Migration in even the oldest cores is no real problem. The recent fuss about “migration” speed was deduced from the Siple core, based on layers where remelting occurred, something not seen in any high elevation ice core like Vostok or Dome C. It remains to be seen to what extent the Siple Dome results are applicable to other ice cores.
But if there was even the slightest migration of CO2, that would affect the ppmv/°C ratio of the above Vostok CO2/temperature graph over time: the proxy temperature indication is fixed in the ice, while CO2 is measured in the gas bubbles. If there was any substantial migration of CO2, the ratio between CO2 and temperature over warm and cold periods would fade away over the recurrent 100,000 years of time difference between the warm periods, but that is not observed.
The ice cores are a reliable source of knowledge of ancient atmospheres, if handled with care. The resolution heavily depends of the accumulation rate, with as result that the data measured in enclosed air bubbles are smoothed, ranging from 8 years for the past 150 years to near 600 years for the past 800,000 years.