Antarctic Ice Cores: The Sample Rate Problem

Prezakly. There is little or no chance that spikes would be preserved. And the baseline error is necessarily downward.

Piers Corbyn says “you aint seen nothing yet”. This guy can save lots of lives and keep us much safer that the current crop of forcasters.

Spector;
Don’t think so, but the bubbles have a thin cold-water “pressure” film, and cold water aggressively dissolves CO2. IMO, that would tend to transport it out of there.

bubbagyro says:

Yes. The experiment, CO2 diffusion in ice, has been performed at Scripps in 2008, demonstrating unequivocally that CO2 does indeed diffuse through ice at a rate that would be significant over years or decades timeframe, much less millenia or megayears, and initial concentrations will decrease if they are higher than the current ambient concentrations:
“CO2 diffusion in polar ice : observations from naturally formed CO2 spikes in the Siple Dome (Antarctica) ice core
Auteur(s) / Author(s)
AHN Jinho (1 2) ; HEADLY Melissa (1) ; WAHLEN Martin (1) ; BROOK Edward J. (2) ; MAYEWSKI Paul A. (3) ; TAYLOR Kendrick C. (4)
Journal of Glaciology ISSN 0022-1430 CODEN JOGLAO
Source / Source
2008, vol. 54, no187, pp. 685-695 [11 page(s) (article)] (1 p.1/2)”

Here’s a copy of that paper, so that we are all on the same page: http://ns.geocraft.com/WVFossils/Reference_Docs/CO2_diffusion_in_polar_ice_2008.pdf Reading that paper does not seem to lead to the conclusions you draw. In particular, they note:
(1) “Smoothing of the CO2 record by diffusion is one to two orders of magnitude smaller than the smoothing by diffusion in the firn at the depth of 287m (gas age = 2.74 kyr BP) in the Siple Dome ice, and so does not degrade the record.”
(2) At depths where the ice age is tens of thousands of years (they use 80,000 years as an example), the diffusion may be comparable to the diffusion in firn and hence it may have “an impact on the smoothing of CO2 records on a decadal scale”. However, they admit that “there are no decadal CO2 data for
ice that is 80 kyr old”. Hence, it would seem that their conclusion about smoothing is, at least at the moment, basically academic.
(3) They also note factors that make the diffusion less of an issue for other ice cores: “Ice cores from colder sites than Siple Dome would experience slower CO2 diffusion in deep ice. The formation of clathrate ice (bubble-free ice) at depths from 500 to 1200m (25–65 kyr) at other Antarctic cold-drilling sites (Vostok, Dome Fuji and EPICA Dome C) is expected to result in highly reduced gas diffusion (Salamatin and others, 1998).”
Perhaps your misreading of the paper stems from the fact that you are reversing how the timescales work. You seem to think their talk about what happens on years to decades mean much more diffusion will happen over longer times. However, what they are doing is looking at 80,000 year old ice and saying, “The amount of diffusion in this ice might be enough for the CO2 to migrate enough to smooth out the CO2 data over a time period of years to a decade or so.” Do you understand the difference?

RE: bubbagyro January 1, 2011 at 6:56 pm –
What this means is that 600 ppm in a bubble in ice with 300 ppm outside will deplete CO2 in the bubble until the outside concentration approaching 300 ppm is asymptotically reached.
But, what about a 300 ppm bubble forming in an atmosphere of 600 ppm? The comment is silent on this point.
Should we expect the behavior to by symetric? There are forces that argue against symetry.
First, there is size and mass of of the CO2 molecule compared to N2 and O2. During compression and closing of the bubble, the molecular sieve effects could be come important in a Physical sense. Second, the Chemical effects of solubility at the ice crystal surface during forming of the ice bubble may enhance or counter act the physical molecular sieve.
Third, what about the snow_to_ice accumulation rates with a 300 ppm bubble in a 600 ppm atmosphere versus a 600 ppm bubble in a 300 ppm atmosphere? Does anyone believe they are the same? Up to a couple of years ago, IPCC would certainly propose much less snow in a 600 ppm environment (dare we say “none”) than a 300 ppm atmosphere. Today, with England snow covered and Ireland’s water pipes bursting, it would not surprise me if they suddenly claim more snow at 600 ppm than at 300 ppm – somehow. I guess the IPCC would say the matter is only academic because according the the ice cores, [CO2] has never been as high as 600 ppm and the physical inability of ice to preserve 600 ppm spikes in a 300 ppm average is unproven. But an experiment showing this inability might be easy and inexpensive to perform.
Fourth, to the suggestion of an experiment that snow should be compressed into ice to analyze the potential diffusion. The compression by mechanical means would tell us very little. It has to be compression by snow accumulation. So you can start off with a steady accumulation of snow, but very the atmosphere in [CO2] over time and see if the forced [CO2] is preserved in the bubbles. Rate of accumulation would be a critical variable: could a 1 yr test represent a 100 year natural accumulation?
While it could be done, more rigorous tests would be to vary the atmosphere above the snow in temperature and pressure as you accumulate snow. In the course of forming an ice bubble (50 to 500 years) the atmospheric pressure possibly changes by at least 20 mbar weekly from weather systems. That would be 2500 to 25000 kneadings of the entire firn gas column as the column is degassing during compression. It is not much of a pressure change, but repeated thousands of times during the formation of the ice bubbles, is it insignificant factor in the bubble’s composition?

This seems to me to be another classic case of experts in one discipline not consulting wide enuf with other disciplines. The issues over at CA demonstrate the folly of not using the correct statistical methods, and O’Donnell et al demonstrate the problems of geophysicists likewise not consulting with statiticians. I was originally a physical chemist, and the issue of ice core veracity has always troubled me.
I guess my question is why don’t these ice core people talk to the physical chemistry people.

This is not explained very well.
Is David proposing that CO2 is being removed from the snow by diffusion but N2 and O2 do not diffuse out of the snow thereby altering the ratio?
I am going to make a wild guess that the relative diffusion rates of gasses in ice are known. So what are they. My intuition says that CO2 would be slower as the molecule is bigger and this is the opposite if what is leopard.

bubbagyro says
———
What this means is that 600 ppm in a bubble in ice with 300 ppm outside will deplete CO2 in the bubble until the outside concentration approaching 300 ppm is asymptotically reached. A concentration of 1200 ppm will diffuse twice as fast as 600, 2400 ppm 4 times as fast, etc. Twice the age, conversely, will have twice the time for diffusion, so old cores will be nearer outside equilibrium level.
—————
this is relevant but it ignores an important factor: the continuously increasing depth of snow. This increases the diffusion length and makes interchange with the atmosphere slower..

Michael [January 1, 2011 at 6:00 pm] says:
Michael [January 1, 2011 at 8:13 pm] says:
Michael [January 1, 2011 at 8:38 pm] says:
Michael [January 1, 2011 at 9:05 pm] says:

OFF-Topic: Hey Michael, are you still celebrating New Years, or are you just testing the moderators?
ON-Topic: Has anyone ever heard of any discoveries of pristine air samples perhaps trapped in a bubble inside an amber sample or some other well-preserved ancient compound? It seems to me that this may actually be possible. I know this is true because I saw it in Jurassic Park.
Another thing I have wondered about is whether anyone has tried to measure CO2 from *new* 5 or 10 year old ice in both the arctic and antarctic. This could be measured by a trusted scientist and then compared to recent CO2 samples from Hawaii. They had better match up or else, Houston, we’ve got a problem.
What do ya’ll think?

While the analysis is very intersting, I have my doubts about the shape of the above results. That is, if this is supposed to be a physical process of diffusion (some form of time-averaging of the CO2 signal).
For some type of averaging process, I’d expect the more heavily averaged signal (blue on the LHS of the above chart) to lie somewhere inside the range of the red ponts.
Instead, the points on the most extreme LHS are below the range on the RHS. And for that to happen, I assume there would need to be a process which removes CO2 from the ice (e.g. by chemical reaction).
The least we could then say would be that the term “diffusion” is confusing.

Michael says
“The markets will front run the collapse. They seem to be starting to do that already. Buy Silver and Gold.”
Sound advice. Note, however, that it is the metal which is necessary. When markets collapse, paper claims to metal will collapse along with all else. Owning real metal is a cumbersome practice, unfortunately.

Don’t forget that the density of the snow/firn/ice increases as you go down, so freshly trapped bolus of CO2 will find it easier to diffuse sideways and up. In those directions most of the diffusion is through the gas phase, while downwards, diffusion is increasingly through the liquid and solid phases. Thus, even if the gradient of CO2 concentration is steeper going down, the rate of diffusion in that direction will be reduced compared to what it would have been if snow density was constant in all directions.

Ah, Dave…I just looked at the Law Dome data ( ftp://ftp.ncdc.noaa.gov/pub/data/paleo/icecore/antarctica/law/law_co2.txt ) and I notice a little bit of a pattern in time…like the only data above 300 ppm is that dated after 1900 and, in fact, the data is not a scattering of points centered around a higher value…It is simply the march of CO2 levels upward in the last 100 years or so. The plot that you made completely obscures the very strong, steady dependence of the CO2 level on the mean air age in the Law Dome data! Since the Taylor core has such a low accumulation rate, it doesn’t have any data points in the post-industrial era!
All that you have shown here is that if you don’t have enough data points to have ANY in the industrial era then you don’t see the post-industrial rise and if you do have any in the industrial era then you see the industrial rise.

bubbagyro says
———–
What this means is that 600 ppm in a bubble in ice with 300 ppm outside will deplete CO2 in the bubble until the outside concentration approaching 300 ppm is asymptotically reached. A concentration of 1200 ppm will diffuse twice as fast as 600, 2400 ppm 4 times as fast, etc. Twice the age, conversely, will have twice the time for diffusion, so old cores will be nearer outside equilibrium level
————-
So are you saying that CO2 in deep ice is in equilibrium with the atmosphere?

Wow, nearly missed this posting, as it was released while the previous discussion of ice cores still is in it’s final state…
To begin with: sample rate indeed has something to do with accumulation, but as the technique of sample analyses evolves, smaller and smaller samples are needed. Thus the relation is not 1:1, but sampling frequency increases for the more recent ice cores. That makes that:
This makes it very clear that the low CO2 values in the Antarctic ice cores during the Holocene could easily be the result of diffusion and do not constitute valid evidence of a stable pre-industrial atmospheric CO2 level of ~275 ppmv.
Is a little premature, as the correlation between sample rate and/or accumulation rate with CO2 levels is mostly spurious (as good as the ice age – gas age “correction” by J.J. Drake is based on a completely spurious correlation).
What happens in the ice core with CO2 during the firn densification was measured in the Law Dome ice cores by Etheridge e.a.:
http://www.agu.org/pubs/crossref/1996/95JD03410.shtml
unfortunately behind a paywall.
Migration of all molecules takes place in all directions, faster in the upper part of the firn, slower with depth, as the pores become smaller. That makes that:
1. the heavier isotopes and heavier molecules are enriched (1-2%) by graviation with depth. That is corrected for, based on the measured 15N/14N enriching.
2. the gas composition at bubble closing depth is a mix of several years. This averaging depends of the diffusion speed and the years necessary to reach the depth where vertical diffusion stops (mostly slightly above closing depth, due to more dense winter layers).
Below the vertical diffusion stop depth, the compositions doesn’t change anymore and air samples in situ taken from the still open pores at bubble closing depth and from the bubbles in the ice cores (via the normal route of relaxation and transport, crushing under vacuum over a cold trap) have the same CO2 level.
In the case of the Law Dome ice cores, the composition of the air at closing depth was only 10 ppmv CO2 below the CO2 level of the atmosphere. As that happened in the period that direct measurements were already done for decades at the South Pole, one could determine that the average age of the air was about 7 years older than at the surface (with a theoretical averaging based on diffusion speed of about a decade). That makes that the top meters of the ice cores and the direct measurements have an overlap of about 20 years. See the overview here:
http://www.ferdinand-engelbeen.be/klimaat/klim_img/law_dome_overlap.jpg
As one can see, the ice cores CO2 simply follows the atmospheric CO2 levels, be it smoothed over several years and is somewhat older (until much older), depending of the accumulation rate.
About diffusion during densification in the firn:
As already said by others, diffusion happens always from the highest level to the lowest level (in fact both ways, but faster from the highest level than reverse), as long as the pores are wide enough. Thus it is impossible to find 290 ppmv in the ice cores, if the atmosphere was higher than that for a period longer than the gas age averaging. Thus if one measures 290 ppmv in an ice core, either that was the average of a quite stable period with 290 ppmv in the atmosphere or it is the average of (large) swings over and below 290 ppmv. Thus any proxy that shows higher or lower levels over a period long enough to be detected in the ice core(s) averaging is biased. That includes historical measurements over land and stomata data…
The (theoretical) smoothing of a peak of CO2, if that happens in a short period, can be seen here for the Law Dome ice core, together with a lot of other interesting items:
http://courses.washington.edu/proxies/GHG.pdf

As Tim Folkerts already said, the diffusion (theoretically) calculated from migration in the Siple ice core is very small. Here a comment on the same point I have made on the other ice core discussion:
The migration of CO2 in ice cores was theoretically calculated by looking at the migration in remelted layers of a “warm” (-24°C) ice core, Siple Dome:
http://catalogue.nla.gov.au/Record/3773250
The extra smoothing of the gas age averaging caused by migration is about 2 orders of magnitude less than the averaging itself. For the Siple Dome ice core, the average gas resolution is about 22 years. Migration increases that with about 0.2 years.
At lowest depth, the layers become thinner and migration may relatively increase and give a doubling of the resolution to some 40 years.
The Vostok and Dome C ice cores are much older at depth, but also much colder (-40°C), which means less water (no water layer at all at -32°C, as long as no salt inclusions are involved). But there is a simple proof that migration doesn’t play a role at Vostok and Dome C:
The Vostok (and Dome C) ice cores show a quite nice relationship between the temperature proxy (dD and d18O) and CO2 levels, be it with a lag. Nevertheless, the ratio between CO2 and temperature is about 8 ppmv/°C between interglacials and glacial periods over the past 420,000 years (recently confirmed for the full 800,000 years in Dome C). If there was some migration, the ratio would fade over time for each 100,000 years period further back. That is not the case. Thus there is little migration of CO2 in the Vostok and Dome C ice cores…

fhsiv says:
January 1, 2011 at 6:46 pm
Is the pattern of CO2 concentration versus depth/age the same in other ice cores? Does the CO2 level out at the same concentration? Is this what the warmers are relying on as evidence of a lower and stable pre-industrial CO2 concentration?
There are huge differences with depth, simply because there are huge differences in accumulation rate. Despite that, it is possible to calculate (and measure for the high accumulation ice cores) the average age of the air at closing depth. Gas composition from extreme different ice cores (temperature, salt/dust inclusions) is the same (within 5 ppmv) for the same gas age in overlapping periods. Here for the past 1,000 years:
http://www.ferdinand-engelbeen.be/klimaat/klim_img/antarctic_cores_001kyr.jpg
Other processes play little role in Antarctica (more in Greenland ice cores): some bacteria use CO2 for DNA repair to survive the Vostok temperatures (0.1 ppmv!), clathrate formation plays a role, but clathrates are effectively destroyed under vacuum (at measurement time)…

Terry says:
January 2, 2011 at 2:04 am
I guess my question is why don’t these ice core people talk to the physical chemistry people.
They did and they do. Here a good introduction in ice core physics and (isotopic) chemistry:
http://www.pnas.org/content/94/16/8343.full

latitude says:
January 2, 2011 at 2:56 pm
I’ve wondered how accurate those ice core Co2 levels really are…..
====================================================
Microbes can survive ‘deep freeze’ for 100,000 years
They can “survive” but that is pretty all they can. In the Vostok ice core they use CO2 for DNA repair. The total amount used, when calculated from the total N2O level, is about 0.1 ppmv CO2. See:
http://www.pnas.org/content/101/13/4631.full.pdf+html item K.

As Beng says above (JAN 2 9:30AM) the peaks and troughs in the ice core record corresponding to the interglacials and glacials have about the same values over an approximately 800,000 year period.
Why?
If significant diffusion were persisting over the whole of that period, those peaks would all be diminishing over time.

Note the van Hoof references
and : .

Given that we know weather and therefore climate is a dynamic process and that dynamic processes tend to have relatively high frequency signals of what ever it is we are measuring. I am the least bit surprised. I am also less then impressed with the ideas surrounding these space limited samples such as ice cores or tree rings. I am not convinced they represent what many seem to think they do. Finer the measurements the less reliable the representativeness of large districts or areas, let along the entire planet.

latitude says:
January 2, 2011 at 4:45 pm
and over the course of a few thousand years…………….You’re diffused out
If there is little difference between the levels, there is no driving force to diffuse out to another. If the microbes use or produce anything, there is a driving forse, but the diffusion rate decreases with the square of the distance and the reciproke of temperature. We are talking about microns between bacterial encapsulating and air bubbles, not meters of ice…
latitude says:
January 2, 2011 at 6:37 pm
Saying they would survive and that’s about all, is not true. They would slow down, but they are still performing functions. Something they might do in a day at normal temperatures, might take a decade or a hundred years or a thousand years, but it will still happen.
Please read the reference of bacterial functions in ice I gave you: the conditions in the Vostok ice core are such that the bacteria’s only remaining function is DNA/cell repair at an extremely low rate, all other functions are so slow as undetectable.
From that source:
Nitrifying bacteria with metabolic rate 10^-12 per hour at -40°C (point L) have been encased in liquid veins in Vostok ice for ~140,000 yr. At a rate of 10^-12 per hour, it takes 10^8 yr to turn over the carbon in their cells without growth.
That is 100 million years to replace all carbon. One million years to replace 1/100th of the carbon content of the bacteria to survive. I have no idea of the total carbon content of all the microbes encapsulated in the Vostok ice cores, but I suppose that that it is far less than the CO2 content…
Methane and other gases only have a significant migration rate (relevant for past air composition) at the lowest few hundred meters of the Vostok ice core (as well as in other ice cores – Greenland), where the temperature is increased due to earth warmth. That part of the core is not used for air measurements, also because the layers are disturbed. Mid-latitude glacier ice cores with higher average temperatures show relevant biolife and Greenland ice cores are unsuitable for CO2 levels, due to frequent (acid) volcanic dust inclusions from nearby Iceland.

Geoff says:
January 3, 2011 at 12:40 am
One more time:
van Hoof references:
From the first link:
The magnitude of the observed CO2 variability implies that inferred changes in CO2 radiative forcing are of a similar magnitude as variations ascribed to other forcing mechanisms (e.g. solar forcing and volcanism), therefore challenging the IPCC concept of CO2 as an insignificant preindustrial climate forcing factor.
As said before in the other ice core/stomata discussion, stomata data are based on CO2 levels as being available over land (during the previous growing season). These are far more variable than and (positively) biased compared to “background” CO2 levels measured in 95% of the atmosphere (including smoothed in ice cores). Even if the bias over the past century can be removed by calibrating against ice cores, that is no guarantee that the same bias was at work in ancient times (changes in land use, climate related vegetation changes, climate related changes in main wind direction,…).
Further, a local increase in the first 1,000 meters to 1,000 ppmv CO2 over land has hardly an influence on IR retainment and thus temperature: less than 0.1°C. Thus the influence of +34 ppmv near ground has no detectable influence at all.
The remaining question is in how far the local measured CO2 levels over land via stomata data can be used as reference for global CO2 levels…

There are several things that do not seem justified in this analysis.
1/ You seem to be of the impression that “diffusion” of CO2 in the ice can somehow account for its concentration being lower than it appears to be in order to suggest a higher historical level and thus less human emissions. Where do you think it could “diffuse” to ?
All this can explain is the inability to capture the high frequency signal but that is nothing new and widely accepted.
2/ Figure 1 has Vostok in the title. That seems to be completely erroneous since there does not appear to be any Vostok data show, just Law and Taylor.
3/ You give no reason or justification for you idea of fitting an exponential to the two sets of data. Why not a straight line or a square law or a hyperbola ? It seems an arbitrary choice.
4/ Why do you do your fit to an ensemble of the two datasets. You are implicitly assuming parallel physical processes at the two sites. This assumption is not explained or justified and in particular does not seem reasonable. Even an eyeball glance at the two sets of data does not suggest that they are two parts of the same thing. Try fitting an exponential to each data set in turn and they will be wildly different.

This makes it very clear that the low CO2 values in the Antarctic ice cores during the Holocene could easily be the result of diffusion and do not constitute valid evidence of a stable pre-industrial atmospheric CO2 level of ~275 ppmv.

All I can read in this data “correlation” is a general trend to increasing CO2 levels that vaguely matches the increasing density of the ice with depth. These are two totally unrelated physical effects with a vaguely similar trends. Neither of these trends is a discovery and any correlation is neither surprising nor useful.
Sorry, but this examination makes nothing “very clear” apart from the prejudice you brought to the exercise.
It appears beyond all else that you are doing what most of climate science seems to be doing of late: starting out with a preconceived idea of what is the truth and setting out to prove it. Your prejudice is the opposite but the method is essentially the same.
The major difference is that you seem to be untrained and hence mistakes are excusable. Those with PhDs to wave about can not plead ignorance and are intentionally misleading the world.

atmospheric concentrations of the past are the only major values which have originated numerous hypotheses concerning the past of our Globe, based on a unique measure : ice cores. All the theories concerning very different scientific subjects have been routinely (and successfully) confirmed by different measures using totally different properties.
So, it should be not at all surprising to discover one day or the other that this beautiful house of cards has suddenly collapsed due to new, simple observations.

Geoff Sherrington says:
January 3, 2011 at 2:55 am
Ferdinand, what is causing what? I can see no logical closed loop explanation of the quote above. English is my first language, but we must speak physics also. For example, some argue that TSI has hardly varied (Leif) and others note that the gas and particle composition from one volcano to another is not constant, thus allowing more unknowns in reconstructions.
Leif is right about the TSI, but TSI is not the only (or even main) influence of the sun (see the latest discussion about the changes in UV). And indeed one volcanic explosion is not another… Anyway, what Van Hoof suggests is that if there was a higher variability of CO2 in the past, then the influence of CO2 on temperature/climate was higher than expected. I doubt that: the variability of CO2 in the stomata data anyway is local/regional variability with (very) limited influence on climate. Only if the local variability really reflects a global variability, then it may be influencing climate to some extent.
My impression is opposite: a small change in temperature causes a small global change in CO2 levels, but may induce a more important change in local/global CO2 levels over land, as both soil bacteria become more active and plants grow harder (if not beyond their optimum). The result is an increased positive bias, as the effect of night respiration on CO2 levels increases more than daylight photosynthesis, as during the day there is a better mixing with the overlying (background CO2 level) air. Many other factors may be involved too: During the LIA, the Gulf Stream was far more South (Portugal, later North Africa) than during the MWP or current. With possible changes in main wind direction (and thus average CO2 levels over land)…
Also, still unanswered, what was the climate like at the time corresponding to the deepest ice in Vostok? Or anywhere else that has been measured deep in the Antarctic?
There are several indications of the climate of many millennia ago, mainly based on isotope changes:
– dD (deuterium/hydrogen ratio) and d18O in the ice water molecules reflect the temperature of the oceans where the water evaporated and partly the temperature of the air where the vapour cooled down to water/snow. See Jouzel e.a.:
http://parrenin.frederic.free.fr/PRO/publications/download/articles/jouzel-JGR2003.pdf
For coastal ice cores, the precipitation comes mainly from the nearby Southern Ocean, while the high altitude inland cores reflect more the whole SH oceans temperatures.
– d13C of CO2 in the gas phase reflects increasing/decreasing vegetation.
– d18O from N2O in the gas phase is inversely correlated with ice sheet formation (I don’t remember why…)
etc…
For Greenland, there are papers speculating that ice core isotopes reflect, in part, the direction of winds bringing snow to the accumulation point. How is this effect accounted for at the South Pole?
Which is important for Greenland, as there are both lots of land and oceans nearby, but less important for Antarctica, as the circumpolar vortex mixes it all from over the oceans…
Still unanswered, does deep ice from near the South Pole indeed contain fragments of prior ice blown by wind, or does the whole reconstruction depend on the assumption of “pure as wind-blown snow?” Most of the oceans around the South Pole are more than 2,000 km distant.
The few mm ice equivalent per year at the high altitude Antarctic cores indeed may have been mixed or blown away or piled up by winds over several years. Not very important, as the (gas) averaging at Vostok is about 600 years… The ice layers are not counted as yearly layers either. Ice samples were taken each meter, reflecting about 20 year intervals (I suppose that this reflects the average years of the sample too) at the top to over 600 years at lowest depth.
Still unanswered for J.J. Drake, how does one assume a conventional correlation of time with isotopes when the gas/ice age difference can be from 2,000 to 6,500 years?
gas age and ice age are independent of each other. There is a direct dependence of ice age at closing depth with the accumulation rate and the closing depth itself depends of temperature. The gas age at closing depth depends on migration speed, which depends on pore diameter (=ice density) and temperature. Thus while there are common factors at work (accumulation rate and temperature influences the period that the pores still are open enough to allow migration) there is no physical connection between ice age and gas age at all.
The main problem is that ice age is easy to establish: directly via counting the layers or indirectly via different methods: conductivity, H2O2 levels,… But gas age isn’t easy. There are gas diffusion models, which are confirmed by direct measurements in shallow ice cores (Law Dome), but an additional problems is that during glacials there is far less precipitation, thus the gas age may get a lot younger, thus increasing the ice age – gas age difference.
Anyway, even within this uncertainty, there is enough evidence that CO2 levels of the same age in the gas phase follow the temperature changes (from isotope changes in the ice phase), not reverse. And that there is a quite nice fixed ratio between CO2 levels and the temperature (proxy). The ice age – gas age difference plays no role at all (as long as the average dates for each phase are correct).
This is getting quite long, more in next message…

Fig 1 seems to just show that more chronologically compact (and therefore deeper and older) ice has less CO2 than less chronologically compact ice. Diffusion before the bubbles are sealed implies that the CO2 record will be smoothed considerably at a decadal or century time scale for most cores, or at a millenial time scale for Vostok, which has very slow accumulation. But it won’t change the average CO2, so the AGW argument that current CO2 (approaching 400 ppm) is substantially higher than the last several thousand years (on average) still stands.
However, Dana Royer and co-authors (GSA Today, March 2004, pp. 4-10 and Geochimica vol. 70, 2006, pp. 5665-75) note that over most of the last 550 million (not thousand) years, CO2 has mostly been in the range 1000-3000 ppm, and that levels under 500 ppm like the last few million years and the late Carboniferious/early Permian period have generally ben periods of glaciation. They characterize under 1000 ppm as “cool”.
This implies that life in general and in particular the oceans do just fine with 1000-3000 ppm. It’s hard to say what the causality is — does high CO2 cause warmth or does a warm climate cause high atmospheric CO2 as the oceans degas — but it’s worth considering whether 500-1000 ppm might protect us against the otherwise inevitable next ice age without excessive warming.
Bubbagyro (1/1 @ 6:56 AM) makes the very interesting point that CO2 may diffuse through ice. This implies that ice absorbs some CO2, and this absorption must vary with temperature. Snow itself will have some CO2 as it falls, but on the ground it will tend to be at a higher temperature than it was when it formed. It may then either absorb or degas CO2 from/into the adjacent air bubbles over time. This could completely screw up the ice core CO2 record, if it is a big enough effect. (Bubbagyro cites Ahn et al, CO2 Diffusion in Polar Ice…, J. Glaciology 54 (2008): 685-95, which I haven’t looked at yet.)
Even if the net absorption/outgassing is small, this diffusion will continue to smooth the CO2 record long after the firn bubbles are sealed, and hence will make the ice core CO2 record look excessively flat.

Geoff Sherrington says:
January 3, 2011 at 2:55 am
More for Geoff:
Re your van Hoof reference, can you point me please to a calibration with instrumentation (such as thermometry) and stomata properties?
and
So, we have some circular argument from where you point us. Using ice bubbles to calibrate stomata to calibrate ice bubbles?
I have no direct references about the influence of temperature, drought, etc… on stomata index level. I suppose these have their influences. My objections against stomata data is that the CO2 levels of which they are based on are not background CO2 levels, more variable and positively biased with unknown changes in local/regional variability and bias over time.
And as stomata are proxies, not direct data, they need to be calibrated: to direct atmospheric measurements when available, or ice cores if not. Not reverse: stomata data accuracy is quite rough (+/- 10 ppmv), compared to e.g. the Law Dome (10-year filtered data) accuracy of +/- 1.2 ppmv. And as you have referenced, they have a limited range: above and below that range, there is no effect anymore. Comparable to the limited response of tree rings to temperature changes…
So, I am not defending the stomata data: they give a good high frequency first approximation, but the ice core data are far more reliable, because that are direct measurements of background CO2 levels, be it smoothed over longer time periods: from a decade to 600 years for the youngest to oldest cores.

Hu McCulloch says:
January 3, 2011 at 6:51 am
Nice to see you here (it is a bit calm at CA these days, I can imagine it that it takes a lot of one’s life to maintain a blog like this one or CA…).
I do agree with most of what you said, but some comment on:
Bubbagyro (1/1 @ 6:56 AM) makes the very interesting point that CO2 may diffuse through ice.
and
Even if the net absorption/outgassing is small, this diffusion will continue to smooth the CO2 record long after the firn bubbles are sealed, and hence will make the ice core CO2 record look excessively flat.
The article shows that the total migration is lengthening the averaging of the CO2 levels by the “warm” Siple ice core from 22 years to 22.2 years at medium depth, up to a doubling (40 years) at deepest ice. The much colder Vostok ice core doesn’t show any flattening of the ratio between (gas phase) CO2 vs. the (ice phase) temperature proxies over each 100,000 interglacial/glacial transition. Thus there was no measurable CO2 migration.

latitude says:
January 3, 2011 at 7:12 am
….
The bigger question is why or what has made CO2 levels so low and are we in danger of CO2 levels being too low.

I’m thinking that in the absence of major vulcanism caused by colliding continental plates to replenish the CO2, the biosphere slowly slurps the CO2 up, converting it into fossil fuels and limestone. Eventually (as during the Carboniferous and Neogene), CO2 gets so low that the globe is vulnerable to ice ages. Perhaps restoring some of that fossil CO2 to the atmosphere (up to 500-1000 ppm) would prolong the Holocene indefinitely?
Just a suggestion, but no crazier than CAGW…

“”””” Joel Shore says:
January 1, 2011 at 7:48 pm
Ah, Dave…I just looked at the Law Dome data “””””
Joel, if there did exist such an out diffusion process; as has been postulated here, and evidently confirmed at Scripps, would not that lead to a situation where the very newest core samples would be expected to be most like the conditions at formation; and older samples would be expected to gradually fall further back from the formation conditions.
So the signature for such an effect would be a rising CO2 (sample) record for the last century or even a couple of centuries.
So your argument that the rising data in the cores since 1900 represents the rising atmospheric levels (which we know confidently from Mauna Loa only since 1957/8) , is not really any more robust than the model where the trapped gases simply out diffuse per Fick’s Law.
As an aside; in the early days of “Sampling Oscilloscope” technology; that being basically the early 1960s for commercially available instruments, it was believed that the rise time of the “scope” was limited by the length of the sampling pulse; so designers (at Tektronix, and Hewlett Packard) spent time trying to make ever shorter sampling gate drive pulses; using Tunnel Diodes, and the like. These efforts were hindered by the ability to make stable very low capacitance sample storage capacitors, so only small amounts of charge were collected during the short sampling pulse; so the early instruments were quite noisy; the noise being of the order of the square root of the number of electrons collected. (Shot Noise )
In the later 1960s it was discovered that in fact the rise time (inverse bandwidth) was not set by the width of the sampling pulse; but was instead determined by the time taken to switch the sampling gate off; in other words the fall time of the sampling gate drive signal. Short shutoff times were generally easier to achieve, than short pulse lengths.
This is pertinent to the ice core record; as the same phenomenon is in effect. The time taken to capture the atmospheric gas sample is not particularly significant; but the gate shutoff time is; that is the period where the air bubbles and the snow sample is being compacted to a degree, where the sample can be said to be truly encapsulated. And that period is evidently decades to perhaps centuries.
So of course fast spikes to higher CO2 levels are simply not going to be recorded, when it takes so long to close the barn door.
I’m generally skeptical of the value of the absolute levels of CO2 or Oxygen in the ice cores. The O2 in the ice is of no interest, since nobody knows when that H2O molecule was formed; it could be billions of years old; so only the O2 in the atmospheric samples has any proxy value.
But I’m not too concerned about the problem; because I don’t think CO2 is in much of a controlling position, as regards the global mean Temperature; compared to the negative feedback cooling effect due to clouds.

latitude says:
January 3, 2011 at 8:48 am
No I’m not….
“Microbes” have been found in all depths, in all ice cores.
Indeed, but at some depths more than other and some cores more than others. From the link you have read:
Sowers (10) reported peaks in the concentrations of N2O and d15N of N2O and dips in the values of d18O of N2O occurring at the same depths in Vostok ice where Abyzov et al. (1) found peaks in bacterial concentration and where Petit et al. (48) found peaks in dust concentration. The ice temperature at that depth is -40°C. The remarkable correlation strongly suggests that, during the penultimate glacial maximum 140,000 years ago, dust and microorganisms were windborne with greater than normal efficiency and deposited onto Antarctic ice.
The effect of this peak in dust and bacteria is a deficit of 0.1 ppmv CO2 over 140,000 years. If we may assume that the number of bacteria is directly related to dust deposits in the ice core, then the effect over time for the deepest useable part at a glacial maximum in the core is not more than 0.3 ppmv. Hardly a reason to reject the data.
The temperature data from the Vostok borehole should be here:
http://www.nature.com/nature/journal/v381/n6584/abs/381684a0.html
but I don’t want to pay the scandalous sum they ask for an over 10 years old article. An average of -40°C is often said, with the not used ultimate bottom layers above Lake Vostok at higher temperature.
Further, Antarctic coastal ice cores at higher temperature and far more dust/bacteria/salt/algue inclusions show very similar CO2, CH4, N2O,… levels for the same gas age. Thus not much influence from the microbes…

George E. Smith says:
January 3, 2011 at 10:04 am
Joel, if there did exist such an out diffusion process; as has been postulated here, and evidently confirmed at Scripps, would not that lead to a situation where the very newest core samples would be expected to be most like the conditions at formation; and older samples would be expected to gradually fall further back from the formation conditions.
So the signature for such an effect would be a rising CO2 (sample) record for the last century or even a couple of centuries.
Sorry George, but Joel is right: Fick’s Law and diffusion works only from high to low. The only possibility for the opposite flow is by reverse osmosis, which needs a lot of pressure difference (and the right porosity) which doesn’t exist between ice layers at 70 and 71 meter or 3000 and 3001 meter. Thus forget “out” diffusion. It doesn’t exist.
Moreover, if there was out diffusion, that would work for every gas, even better for the smaller ones: methane, Ar, O2, N2… than for CO2. Thus enriching the CO2 levels.
Further, I can only agree with Joel that the sampling rate and CO2 levels have nothing to do with each other: The oldest Siple Dome samples (Neftel, 1992) used two sampling methods: 3 samples, each covering 220 years (!) and a “newer” method, with 12 samples each covering 11-12 years. If you average the higher resolution samples, you will find the same CO2 levels for the same average gas age. See Neftel e.a.:
http://www.biokurs.de/treibhaus/180CO2/neftel82-85.pdf
Further the 3 youngest gas age samples overlap with the direct measurements in the atmosphere and all samples overlap with the Law Dome ice cores samplings, which have a sampling rate of 2-5 years in the overlapping years with similar CO2 levels…
Thus George (and David) there is no Sample Rate Problem, only a problem of a spurious correlation…

tallbloke says:
January 3, 2011 at 11:48 am
Ferdinand, thats interesting, because dust was the reason given for the redrawing of the Greenland ice cores so that they matched the Antarctic ice cores…
http://tallbloke.wordpress.com/2010/12/28/tom-van-hoof-historical-co2-records/
Any idea why that might be?
The difference is in the type of dust: dust in Antarctica is mainly sea salt (composed of mostly NaCl but also carbonate and other salts), far more at coastal sites than at altitude far more inland. During glacials, there is far less water vapour/clouds/rain/snow and dust from far away (even sand) can easier reach the inland ice cores. The highest dust levels in the Vostok ice core are measured just before a new interglacial, when temperatures are at its lowest.
In the Greenland ice cores we have the same points (but in general a one order higher amount of dust than in coastal Antarctic cores) + an additional source: frequent volcanic eruptions from Iceland volcanoes. Iceland is at a hot spot and volanic ash is from the deep earth crust, quite toxic and very acidic (HCl, HF). When that is deposited in the layers of the Greenland ice core, that reacts over time with the sea carbonates and produces extra CO2 in situ…
Some references:
http://cat.inist.fr/?aModele=afficheN&cpsidt=3634884
http://europa.agu.org/?view=article&uri=/journals/jc/97JC00163.xml

“”””” Ferdinand Engelbeen says:
January 3, 2011 at 11:56 am
George E. Smith says:
January 3, 2011 at 10:04 am
Joel, if there did exist such an out diffusion process; as has been postulated here, and evidently confirmed at Scripps, would not that lead to a situation where the very newest core samples would be expected to be most like the conditions at formation; and older samples would be expected to gradually fall further back from the formation conditions.
So the signature for such an effect would be a rising CO2 (sample) record for the last century or even a couple of centuries.
Sorry George, but Joel is right: Fick’s Law and diffusion works only from high to low. The only possibility for the opposite flow is by reverse osmosis, which needs a lot of pressure difference (and the right porosity) which doesn’t exist between ice layers at 70 and 71 meter or 3000 and 3001 meter. Thus forget “out” diffusion. It doesn’t exist. “””””
So you don’t like my usage of terms; so let me redefine; For OUT diffusion read IN diffusion; as in CO2 diffusing INTO THE ICE; which just co-incidently is diffusion OUT of the trapped air samples, INTO the cie walls around the air bubble; where rumor has it the CO2 abundance is much lower than in the air; hence the reason why Fick’s Law would be applicable; and my comment stands.
NO I WAS NOT MEANING DIFFUSION OF CO2 FROM THE TRAPPED POCKETS OUT INTO THE AMBIENT ATMOSPHERE.

George E. Smith says:
January 3, 2011 at 1:51 pm
So you don’t like my usage of terms; so let me redefine; For OUT diffusion read IN diffusion; as in CO2 diffusing INTO THE ICE; which just co-incidently is diffusion OUT of the trapped air samples, INTO the cie walls around the air bubble; where rumor has it the CO2 abundance is much lower than in the air; hence the reason why Fick’s Law would be applicable; and my comment stands.
Quite confusing, your use of IN and OUT, I must say…
Well, let’s take it that way:
– I don’t see any reason why the migration into the intercrystalline waterlike layer would start when all bubbles are closed, it would start at the very moment that the first snowflakes are sintering together (as far as it happens). Or if you really mean the ice itself, then already when the snowflakes were formed.
– The ice is itself is not a matrix that allows CO2 to be included. CO2 in many cases is measured in dry air by trapping water vapor over a cold trap: no measurable amount of CO2 is trapped in the matrix (we are not talking here about nanograms of dope!).
– three methods were/are used to measure CO2 from ice cores: the old method was by melting all ice and applying vacuum to remove CO2 from the water phase. That did show a deficit of a few ppmv compared with the usual method: crushing the ice at low temperature under vacuum and measuring CO2 in the air of the opened bubbles, after removing water vapour over a cold trap.
The third method is sublimating all ice under vacuum and cryogenic separation of all components. That gives the same result as for the second method in total CO2, but is used for isotope research, to avoid fractionation.
Thus even if there was some migration into the (intercrystalline) ice structure, it is unmeasurable, as the comparison between the three methods shows. By far not enough to explain the disappearance of a quarter of the total amount out of the bubbles.

David Middleton says:
January 4, 2011 at 1:45 am
1 meter ~ 1 year at Law Dome (late 19th to 20th centuries).
1 meter ~ 100 years at Vostok (during last Pleistocene glaciation).
1 meter ~ 1,000 years at Taylor Dome (during last Pleistocene glaciation).
3-5 ppmv/m at Vostok is about 3-5 ppmv/century. 3-5 ppmv/m at Taylor Dome can be 3-5 ppmv/millenium. 3-5 ppmv at Law Dome is about 3-5 ppmv/yr.
First, sorry David that I have made a mistake by misinterpreting your “sample” rate, as literal sampling, not m/years…
The interesting point is not the differences between different ice cores, but the changes in sampling rate within one core: Taylor Dome shows a sample rate starting as 10 years/meter and ends with 2,500 years/meter after 230,000 years at 550 meter depth.
Vostok starts with 20 years/meter and ends with 540 years/meter at 3310 meter depth.
Quite huge changes in sample rate with depth and quite huge differences between different ice cores.
The Vostok ice core goes back 420,000 years with ups and downs of 100 ppmv in four cycles, with decreasing sample rate with depth. The interesting point is that there is a quite good correlation between the (ice based) temperature proxy (dD and d18O) and the (gas based) CO2 levels over the four cycles. If there was even the slightest diffusion, that would be visible over time as a fading away of the ratio. As an extra, that would be accellerated by the sample rate, as any diffusion would have the same speed over smaller layers (at the same temperature), representing much broader time periods. This is not the case at all, the ratio remains the same over 420,000 years (and recently confirmed over 800,000 years).
And I like to see a sr/CO2 plot of the full 420,000 Vostok years (have no time now to do it myself)…
Further, if there was diffusion over the first millennia (the Holocene), it would be strange to find such low levels: diffusion averages the levels over time, but it doesn’t change the average. That means that either the real levels at closing time were (much) lower or there was little migration…

Ferdinand Engelbeen says:
January 3, 2011 at 11:17 am
….
The temperature data from the Vostok borehole should be here:
http://www.nature.com/nature/journal/v381/n6584/abs/381684a0.html
but I don’t want to pay the scandalous sum they ask for an over 10 years old article.

The data, at least, is available for free via http://www.ncdc.noaa.gov/paleo/icecore/antarctica/vostok/vostok_data.html .
The assumptions behind the dating are crucial, but unfortunately are only discussed in the article.

Hu McCulloch says:
January 4, 2011 at 8:56 am
The data, at least, is available for free via http://www.ncdc.noaa.gov/paleo/icecore/antarctica/vostok/vostok_data.html .
The assumptions behind the dating are crucial, but unfortunately are only discussed in the article.
Thanks Hu, I have looked there, but the “temperature” I did find was the reconstructed (hemispheric ocean) temperature, based on dD (or d18O). I am looking for the borehole temperature that the Russians have measured from the surface to the bottom of the drilling hole at Vostok (in preparation for an eventual drilling through the last ice to reach Lake Vostok).

David Middleton says:

I’ve now included Vostok back to 157 kya and Taylor Dome (11-27 kya) to the graph…
CO2 v Sample Rate w/ Vostok
When I included the older data, I found that the exponential trend-line was not a good fit; a logarithmic trend-line fit quite well.
Then I plotted the GRIP (Greenland) data (8-40 kya) on the same graph…
CO2 v Sample Rate w/ Vostok and GRIP

But, we already know that your plot is completely devoid of meaning. What you are seeing is that there is a relationship between accumulation rate (which you misname “sample rate”) and what the CO2 level was. However, the reason for that relationship is:
(1) For Law Dome vs. Taylor Dome, you are really just seeing the fact that Law Dome captures some of the modern industrial CO2 rise and Taylor Dome doesn’t.
(2) For the others, you are seeing the fact that accumulation rate for some ice cores is strongly dependent on climate, with the accumulation rate being higher when the temperature is higher because a higher temperature generally results in more precipitation (and temperature and CO2 level are, of course, strongly correlated).
So, your plot doesn’t really tell you anything about what you really want to know. It provides exactly zero evidence for what you claim to be showing evidence of. And, all the other available data (such as the paper that bubbagyro found) seems to imply that the effect of diffusion will range from completely negligible to maybe (in the very worst circumstances) about as large as the effects due to it taking time for the bubbles to close off.

George E. Smith says:

On page 52 of that issue, in that article, Ruddiman shows Carbon Dioxide data from two (Antarctic) ice cores, simply designated as ice core 1 and ice core 2; and they contain data going back for 2500 years. Well ice core 1 goes from 500 BC to about 1600 AD, while ice core 2 starts at 1000 AD and goes up to 1900 AD, so they overlap from 1000 AD up to 1600 AD; so 600 years of common proxy monitoring of atmospheric CO2 abundance (allegedly).
Well the problem is, that during the 600 year overlap period; the two ice core records of atmospheric CO2 are nothing at all like each other and in fact from 1000 AD to 1200 AD the go in exactly the opposite directions, with # 2 rising very steeply and #1 falling at a good clipand then from 1400 AD to a600 AD #1 is rapidly rising, while #2 begins a steep plunge.

George,
Did you notice the scale on the vertical axis? The points from the two cores are **ALL** within 6ppm of each other! The majority are within about 2ppm of each other. Sure, if you are worried about measuring the CO2 concentration to a couple ppm that might be important…but I’ve never assumed that the ice core measurements are precise to that degree anyway! (After all, we know that there are some global and annual variations in CO2 levels on that scale.)

George, my question is already answered by Joel…

“”””” Ferdinand Engelbeen says:
January 4, 2011 at 3:15 pm
George, my question is already answered by Joel… “””””
Ferdinand, I’ll take a look at what Joel had to say regarding your question.
As to the graphs Ruddiman presented in SA, don’t waste your $8.
The actual P-P CO2 range of the two ice cores over the 2500 years; with 600 years of overlap , was from a low of 275 ppm for ice core # 2 at the 1600 end point of core #1 going to about 288ppm at 1900 for the same core.
The #1 core ranged from 277 low 650 BC to a high of around 285 at 750 AD then dropped to 278 again around 1400 to top out at 283 at the 1600 end point.
So maybe as you say; not outside the expected noise levels of those ice cores.
But if so the question remains; where does Ruyddiman get off presneting this data to Scientific American, and its audience if it is all in fact phony baloney.
If I take your caution; which I pay a lot of heed to; there is nothing whatsoever in ruddiman’s graph that he should even be writing about. He even attaches all knids of historical significance to events; whcih may be nothing more than noise, such as blaming the black death on what those ice cores “were doing”around 1350.
Turns out that that is exactly the time period, when core #1 and core #2 tell their most totally different stories.
And a Roman era plague that killed 25-40% of Europeans circa 570 AD corresponds to a wild vertical transient for core #1, while amazingly the black death of 1350 happens at the low of core #1.
Evidently total BS of the highest order; but Ruddiman presents it as gospel truth for SA to swallow. I’m taking them at their word, that the december 2010 issue I received is the last one they will ever send me.
Oh for an encore, Ruddiman takes the very first data point for core #1 at 750 BC and draws a straight line to the end point of core #2 at 1900 and calls that straight line the long term deforestation trend. Only three of the several dozen data points happen to lie above that line, and those three just barely so. Of course the two end points are by design exactly on that line.

George E. Smith says:

As to the graphs Ruddiman presented in SA, don’t waste your $8.
…
Oh for an encore, Ruddiman takes the very first data point for core #1 at 750 BC and draws a straight line to the end point of core #2 at 1900 and calls that straight line the long term deforestation trend. Only three of the several dozen data points happen to lie above that line, and those three just barely so. Of course the two end points are by design exactly on that line.

George,
I agree with you that that sidebar in the Scientific American article is very weak…and some of the claims (like a drop in CO2 around 500 AD) are just weirdly wrong. To be fair though, that is not the most central part of Ruddiman’s hypothesis and it also seems like a part that was messed up quite a bit in the translation between his original peer-reviewed article and the Sci Am popularization. Here is the original one: http://bill.srnr.arizona.edu/classes/182h/Climate/Ruddiman.pdf Note that Fig. 7 is the corresponding figure and it and the discussion of it are considerably better than the Sci Am article (e.g., he tells you where the data is from, he is more clear about the timing of various plagues etc, …) Probably the most important clarification is that the “deforestation trend” line that you objected to is clearly explained as being from other longer term data not shown, not simply from connecting the two points at the two ends of this graph. (See Fig. 2 for the data that was presumably used.)
Anyway, I am not really trying to defend Ruddiman’s hypothesis (which I have no strong opinion either way on), and especially this particular aspect of it which really seems to be pushing the ice core data beyond its reliabability…but just wanted to point out that going to the original peer-reviewed source you do at least get a better picture of what he is trying to say.