In climate skeptics circles, there is rather much confusion about historical/present CO2 measurements. This is in part based on the fact that historical direct measurements of CO2 in the atmosphere by chemical methods show much higher values in certain periods of time (especially around 1942), than the around 280 ppmv which is measured in Antarctic ice cores. 280 +/- 10 ppmv is assumed to be the pre-industrial amount of CO2 in the atmosphere by the scientific community. This is quite important, as if there were (much) higher levels of CO2 in the recent past, that may indicate that current CO2 levels are not from the use of fossil fuels, but a natural fluctuation and hence its influence on temperature is subject to (huge) natural fluctuations too and the current warmer climate is not caused by the use of fossil fuels.

As we have looked at the modern measurements here,  there is little doubt that humans are responsible for the increase of CO2 in the past 50 years, and probably over the past 150 years. But what about the historical measurements? Shouldn't these not be taken into account? Was Callendar so wrong by discarding outliers, based on stringent criteria? That is the thesis of Ernst Beck [1], who compiled thousands (90.000!) of historical data from all over the world, averaged them and found a correlation between the average historical CO2 levels and temperature in the century before the Mauna Loa (and other station) data came into play.

But there are a lot of problems with the historical data. Many of these problems make that the historical data have no relationship with global CO2 levels at all. Therefore we will have a look at the historical data with the current knowledge of CO2 measurements in mind. The following is a step-by-step report of the essential points in the many personal discussions I had with Ernst Beck. Follow the steps:
  1. Accuracy of historical CO2 measurement methods.
  2. Variations due to local circumstances.
  3. Minimum and average values of the historical data.
  4. The probability of a huge peak around 1943.
  5. Conclusion
  6. References

1. Accuracy of historical CO2 measurement methods:

Historical measurements of CO2 in the atmosphere were based on chemical methods. The accuracy was quite good for several of the methods, including CO2 in air, but less appropriate for atmospheric CO2 measurements for other methods.
Many of the methods were accurate to 3% of the measurement, thus for data around 300 ppmv, that is an error of app. +/- 9 ppmv. For that time a good performance.
One exception is the micro-Schollander method, which was intended for exhaling air oxygen and CO2 (around 2%)  measurements. This used ambient air for calibration of the equipment. If ambient air was between 200-500 ppmv, the instrument was deemed OK. That is no problem for measuring CO2 in exhaled air, but it is a big problem if you want to know CO2 at levels of around 300 ppmv... While the place of investigation (Barrow, AK, USA) was of particular interest (one of the current base stations for CO2 measurements is there), the performance of the equipment doesn't allow for any conclusion.

For many of the historical measurements, there is little description of methods used, sampling, sample handling, calibration, etc... Thus while a certain method may have been sufficiently accurate, the sampling/handling and/or lack of calibration,... may be the origin of large changes in performance and/or huge changes in data obtained. That is also what a comment by peterd on a blog [7] says, a person who also looked at historical data, but beyond what Ernst Beck investigated.
From that source:

Caldwell performed five series of tests comparing the Pettenkofer method with known values of CO2 and with the Letts and Blake modification of the Pettenkofer, which was itself of a high accuracy when compared with known CO2 volumes. His summaries show actual CO2 concentration to vary from 0.66 to 0.89 of the amount measured by the Pettenkofer method.”
In other words, the Pettenkofer values- and, by implication, many of those reported by Beck as supporting his >400 ppm values in the 1930s-1940s- may have been over-estimated by 50%!

Some of the historical writings only give an average of the measurements, some only have one or a few measurements. In general, where large measuring series are involved, the equipment and chemicals used were more frequently quality controlled than in short series. But that is not a guarantee for accuracy or relevancy of the data.

2. Variations due to local circumstances:

This is the main problem with the historical CO2 measurements: Near all data come from sites on land, some in towns, some in (rice) fields, some in forests and grassland, or a mix of them in the main wind direction (or different for different wind directions). And near all are measured near ground 0-1 m high. For the oversight of historical measurements, it is important to know where and how was sampled and measured, including wind direction and surroundings. In this chapter we will try to compare modern measurements over land with the historical measurements, and with background measurements.

As is clear for modern measurements, the CO2 levels measured at low height over land don't give a clue about the "background" level of CO2. Therefore the influence of local sources/sinks is huge and give a (mostly positive) bias to the measurements. Here an example.

Diekirch (Luxemburg) [2], has a weather station is in a valley with forests, urbanisation and traffic in the main upwind direction. The main air intake is at about 20 m above the valley floor. The diurnal CO2 level change is heavily correlated with the buildup of an inversion layer at night, which is anti-correlated with temperatures, sunlight and wind speed. At windspeeds > 1 m/s the inversion layer is destroyed and the diurnal variation is much smaller than at low wind speeds.
Giessen in Germany [3] had a weather station which made a lot of measurements in 1939-1941 at different heights: zero, 0.5, 2 and 14 m. Some comparison here is possible.

diurnal CO2 Diekirch
Diurnal CO2 measurements compared to temperature, wind speed and sunlight in Diekirch, graph from [2]

diurnal CO2 Giessen
Diurnal CO2 measurements at Giessen, taken from [3]

The data from Giessen have an average of 462 ppmv (data in the paper are expressed in thousandth of a percent, which is 10 ppmv), with a minimum of 240 ppmv and a maximum of 680 ppmv. Variability is +/- 132 ppmv (2 sigma) at 14 m. The background level in that period was about 310 ppmv (Law Dome ice core).
Diekirch has an average of 405 ppmv and a variability of +/- 30 ppmv (sigma level not given). The background level in the time period of measurements was 376 ppmv (Mauna Loa).
From these figures, it is clear that the Giessen data have a higher average and larger variability than modern data even from a shielded valley like Diekirch. In both cases, the background level is within the variability around the average. And in both cases, the CO2 levels reach the background level at high(er) wind speeds. That points to the influence of insufficient mixing at low wind speeds and at night when there is an inversion layer.

Update: We recently received the CO2 data from Linden-Giessen 1995-2008 from the Hessisches Landesambt für Umwelt und Geologie (HLUG) [11]. This gives an interesting view on the half-hour CO2 measurements over a day in Giessen, if compared to the hourly averages of stations measuring at places far away from local/regional sources/sinks:

Giessen vs. background
CO2 levels at the south pole, Mauna Loa, Barrow and Giessen (data of Giessen from [11])
for the same days as for Diekirch in the previous figure.
The data from SPO, MLO, BRW are the raw hourly averages, including outliers caused by
volcanic degassing or upslope winds (MLO) or mechanical problems (SPO)

It may be clear from this graph that measurements near ground over land show a huge diurnal variation, while measurements in the "background" atmosphere, far from local sources/sinks, are near equal from near the north pole (BRW) to the south pole (SPO). Moreover, simply averaging the data from Giessen gives a positive bias, compared to the more "global" measurements. This bias is even more pronounced in the historical Giessen data, because of timing of the 3 samples/day taken at 7 am, 2 pm and 9 pm:

Giessen 14 July, 2005
Diurnal CO2 variation of 14 July at Giessen. Data from [11].

The sampling at 2 pm has a negative bias of about -15 ppmv compared to background (380 ppmv), while the samplings at 7 am and 9 pm have a positive bias of about +50 ppmv compared to background for this day. This gives an about +40 ppmv positive bias if one averages the historical data from Giessen. But that isn't even constant, and may differ from day to day and with the seasons, as the CO2 levels and the flanks of the diurnal variation may shift with wind speed, temperature and sunlight...

The influence of wind speed
is obvious in Diekirch:

Diekirch (Luxemburg) CO2 measurements compared to wind speed, from [2].
and obvious in Giessen (1939-1941):
Kreutz influence of wind speed Giessen
Historical measurements of Giessen (Germany), influence of solar, temperature,
wind speed and precipitation on CO2 levels, from [3]

and obvious in the data collection by Misra (1941-1943):
Misra CO2 India
Historical measurements on bare ground and crop fields in India by Misra, from [5]

As can be seen in the Diekirch data, there is an inverse correlation between CO2 levels and wind speed. With higher wind speeds, CO2 levels are better mixed with higher air layers which have "background" CO2 content. This reduces the CO2 content at ground level. The assymptote of CO2 levels at high speed winds (as was seen during storm Franz, 11 January 2007) is about 385 ppmv, very close to the 382 ppmv level found at Mauna Loa in the same period. With low wind speeds, the difference between day and night can reach over 100 ppmv in Diekirch, with higher wind speeds the difference is near zero. In the more variable Giessen data, wind speed seems to have little influence on average CO2 levels, but the above graph shows that low wind speed at night correspond with high CO2 levels. The influence of wind on CO2 levels again is clear in India, even where is measured under the leaves of rice and other crops.

The average of the measurements in Diekirch is about 30 ppmv higher than of the measurements at Mauna Loa (or any other NH station) over the same period. This is an important point, as it makes it clear that simple averaging of land based data, like Ernst Beck has done, doesn't reflect background levels. Something similar is the case as shown for Giessen (and other historic data over land), because the measured CO2 levels have a very wide spread, the sampling is often in the downgoing/upgoing flanks and the mimimum includes the background level measured in the ice cores.

More evidence for near-ground  higher variability and higher average than background CO2 levels can be found in the data of Cabauw (mid Netherlands): they have air intakes at 20 m , 60 m, 120 m and 200 m height. Here a typical plot of 24 hours (12-13 October 2008) and several weeks (14 September - 11 October 2008) of CO2 measurements:

cabauw day - week data
One day and several weeks of typical CO2 measurements from different heights at Cabauw (Netherlands), from [4]

This gives a nice view of the problems with measurements over land in the near-ground layers: the nearer to the ground, the more influences of local sources and sinks are visible, where the sources are more prevalent. The sinks (uptake of CO2 by vegetation) happen to work only at daytime, when temperature/wind/turbulence make a better mix with the overlaying air. This gives a positive bias of averages only taken near ground level. Only in the afternoon and/or with strong winds, the CO2 measurements reach background levels. This makes that most of the historical measurements taken into consideration by Ernst Beck are of no value for the estimation of the real historical background CO2 levels of that time and certainly that a simple average has a positive bias.
In fact, one should have a look at only the lowest values at high wind speed, as these reflect the background CO2 levels more accurately today and thus probably in the past too.

Theoretically, one can try to find a formula to calculate the background CO2 value on the base of the relationship CO2 measurement - wind speed (as was tried in Diekirch), but the formula is not necessarely the same or similar for different places, due to differences in local/regional sinks and sources in the neighbourhood, local crop rotation and growth, topography and local/regional wind patterns. And one need quite a lot of data at high wind speed (over 4 m/s) to have a reasonable approximation of the background level. Unfortunately the number of historical observations with high wind speed at Giessen are too limited to be of value...

3. Minimum and average values of the historical data:

As has been shown in the previous chapter, the averages of near-ground historical measurements over land show a positive bias over the real background levels of that time. Thus it is more interesting to have a look at the spread of the measurements and especially to have a look at the minimum values, as these are probably measured at high wind speed, when measured CO2 are near background. Then we can compare them with the measurements in ice cores.

For the period of interest, where the 100 ppmv peak value of CO2 is situated (1943) that gives this graph:

Beck CO2 data 1930-1950
Minima and averages of historical CO2 data 1927-1953 as collected from the literature by Ernst Beck, from [6]

Except for fall data measured in Vienna (average minima above the 500 ppmv graph maximum!), all minima are below the ice core data. As the ice core contains smoothed values (average 8 years smoothing), this levels out fast CO2 variations in the atmosphere. But the averages measured over land in the period 1935-1950 (15 years) is about 100 ppmv higher than in the ice core. That proves that the land based measurements show positive biased values, which have no resemblance to the real historical background (which was already obvious in the previous chapter). Current background values over the NH (including selected values at Schauinsland, Germany, at 1,000 m altitude) show only a difference of less than 5 ppmv with the South Pole measurements. And South Pole atmospheric measurements overlap with ice core measurements for a period of about 20 years within the error margins (+/- 1.3 ppmv).

Further, there is a discrepancy between the averages Beck calculated and reality: Beck averaged as a matter of speaking 1,000 datapoints from Giessen with 10 from Philadelphia and 10 from over the oceans, thus giving 100 time more weight to Giessen than to the other two. For a "global" average, this is quite remarkable. The take on this by peterd [7]:

Bray’s Table 3 is a summary of data measured between 1868 and 1956 (ie.., just at the point where Keeling Sr comes on the scene). The MAJORITY of these values indicate CO2 around 310 ppm. The high (>400 ppm) values emphasized by Beck belong to a rather small minority of workers. It is as if, out of 100 people doing the same measurement (though maybe by different methods), 99 of them agreed quite well, but the 100th, who nevertheless produced a data set vastly greater than the other 99, was given preferred status, simply because he produced more numbers. Does this sound reasonable? Does this sound scientific? Yet this is precisely what Beck has done, in according preferred status to the high values. Nowhere does he offer any explanation as to why these high values should be believed.

I only can agree with his words...

4. The probability of a huge peak around 1943.

If you look at the previous graph to the averages, there is a huge peak in CO2 around 1943. That is a very fast one: from the base line in 1936 at around 330 ppmv (historical) to 430 ppmv in 1943. That is 100 ppmv in only 7 years time. The same the other way out: back down to the baseline of 330 ppmv in 1953, again 100 ppmv in only 10 years time. That is a huge amount of CO2 which must come and go from/to another reservoir. Such huge movements aren't seen  in any year of accurate measurements since 1959. The maximum variability in recent times is +/- 1 ppmv/yr.
But let us assume that the figures are right. Well, the release of 210 GtC (= 100 ppmv) in 7 years time is theoretically possible as result of a huge release from volcanoes, (undersea) vents, meteorite impacts, etc... Or burning 1/3th of all vegetation on earth... There is no sign that something like that happened, but it is possible. But the opposite way: that 210 GtC were absorbed in ten years time, either by vegetation (that is one third of all vegetation as extra growth) or oceans, is physically impossible. There simply is no process in the natural world which can absorb such a quantity of CO2 in such a short time. This in fact refutes the probability of such a peak value around 1943.

We have other sources of CO2 measurements or proxies which may give a hint of CO2 variability around 1943. None of these show a specific variability around 1943, which should be present if the atmospheric CO2 content increased and decreased with 30% in such a short period. Have a look:

1. ice cores.
Here an overview of ice core measurements, including the period around 1943. Two of the Law Dome ice cores have a resolution of less than a decade. A peak of 100 ppmv thus should show at least 10 ppmv peak in the ice core, even more in this case, as the peak value was high during 9 years. But it isn't visible, except for a small irregularity of about 1 ppmv, but too early, in 1938. In fact there is a dip around 1943, but that has little relevancy, as the datapoints are within the accuracy of the measurements of 1.2 ppmv (1 sigma) for the same core (5 ppmv between ice cores for the same gas age).

ice core CO2
CO2 levels in ice cores around the 1943 "peak" of historical data [8].
2. Stomata data.
Although stomata data have their own problems (positive bias due to local/regional CO2 sources) quite similar to the historical data, they don't show a peak value around 1943:

stomata data
CO2 levels vs. stomata data calibration in the period 1900-1990 [9]

This is the calibration curve of stomata data vs. CO2 in the atmosphere, firn and ice cores, done for the period 1900-1990. If there had been a peak of 80 ppmv around 1943 (that is at 310 ppmv in the ice cores), the stomata index (SI) should have reacted with an extreme low value at the 310 ppmv point. But nothing special can be observed.

3. Coralline sponges.
Coralline sponges grow in shallow waters in the tropics. The carbonates built into the sponges follows the isotopic composition (d13C or 13C/12C ratio) of CO2 in the surrounding water, without modification. But as shallow waters track the CO2 in the atmosphere, the composition changed over the years since the start of the industrial revolution. If there was a huge release of CO2 in the atmosphere around 1943, this must be seen in the d13C ratio, as such a release from vegetation should give an enormous drop in d13C (about 4 per mil decrease, but flattened by the seasonal exchanges), while such a release from the (deep) oceans should give an important increase in d13C (about 1.6 per mil increase, but flattened by the seasonal exchanges). But there is no visible change in trend around that period:

sponges d13C
d13C evolution in coralline sponges [10]

Again, there is no sign of anything happening around 1943, even if the d13C records have a high resolution of 2-4 years around that period and the accuracy is fine enough to detect an extra addition/uptake of 1 GtC (0.5 ppmv) from/by vegetation or 4 GtC (2 ppmv) from the (deep) oceans.

Thus all we can say is that other (proxy) methods, even with high resolution, don't show any abnormal variation around 1943.

5. Conclusion

While I respect the amount of work done by Beck to look at the historical data, I only can disagree to a large extent with his conclusions. Besides the quality of the measurements themselves, the biggest problem is that most of the data which show a peak around 1943 are taken at places which were completely unsuitable for background measurements. In that way these data are worthless for historical (and current) global background estimates. This is confirmed by other methods which indicate no peak values around 1943. As the minima may approach the real background CO2 level of that time, the fact that the ice core CO2 levels are above the minima is an indication that the ice core data are not far off reality.

6. References

Note: as the website of the late Ernst Beck is defunct and the last update was from 2010, several links don't work anymore.
We try to recover as much historical data as possible, but that will take some time...

[1] Main page of Ernst Beck's thesis about the historical CO2 measurements, "180 Years of atmospheric CO2 Gas Analysis by Chemical Methods":  http://www.biomind.de/realCO2/

[2] Seasonal and Diurnal CO2 Patterns at Diekirch, Luxemburg, a very comprehensive overview of (possible) influences on near-ground CO2 levels: http://meteo.lcd.lu/papers/co2_patterns/co2_patterns.html

[3] English summary of the W. Kreutz paper about the CO2 measurements in Giessen: http://www.biokurs.de/treibhaus/literatur/kreutz/Kreutz_english.pdf. The original German paper can be found on the web page of Ernst Beck under http://www.biomind.de/realCO2/literature/CO2literature1800-1960.pdf from [1].

[4] CO2 and other data measurements from the tall tower at Cabauw (The Netherlands). Online updated day and week overview at http://www.chiotto.org/cabauw.html.

[5] R.K. Misra, "Studies on the Carbon dioxide Factor in the Air and Soil Layers near the ground": http://www.biokurs.de/treibhaus/literatur/misra/misra1941.doc , from the web page of Ernst Beck under http://www.biomind.de/realCO2/literature/CO2literature1800-1960.pdf [1].

[6] Excel file of the historical data by Ernst Beck: http://www.biomind.de/realCO2/literature/CO2rawdata1800-1960.xls obtainable from [1] via http://www.biomind.de/realCO2/data.htm .

[7] Comment of peterd about Beck's data on the blog of Jennifer Marohasy

[8] Ice core data from NOAA: http://www.ncdc.noaa.gov/paleo/icecore/current.html

[9] Ph.D. dissertation of Tom van Hoof: "Coupling between atmospheric CO2 and temperature during the onset of the Little Ice Age": http://igitur-archive.library.uu.nl/dissertations/2004-1214-121238/index.htm

[10] Böhm e.a., "Evidence for preindustrial variations in the marine surface water carbonate system from coralline sponges": http://www.agu.org/pubs/crossref/2002/2001GC000264.shtml

Hessisches Landesambt für Umwelt und Geologie (HLUG) at: http://www.hlug.de, with thanks to Dr. Ludger Grünhage of the University of Giessen, for sending the data. The University of Giessen is where a CO2 open air enriching experiment is going on. The monthly average diurnal CO2 data from that experiment also give a good indication of the daily/seasonal variability of the CO2 levels at Giessen and are available here.

On the net: 25 October, 2008.
Last change: 10 July, 2010.
Temporary change: 22 November, 2022.

To the
family home page

To the climate change page

Zend naar: ferdinand.engelbeen@pandora.be