In climate skeptics circles, there is rather much confusion about historical/present CO2 measurements. This is in part based on the fact that rather accurate 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.

To be sure about my skepticism: I like to see and examine the arguments of both sides of the fence, and I make up my own mind, based on these arguments. I am pretty sure that current climate models underestimate the role of the sun in climate variability and overestimate the role of greenhouse gases and aerosols. But I am as sure that the increase of CO2 in the atmosphere since the start of the industrial revolution is mainly from the use of fossil fuels.

There are several reasons why the hypothesis of large non-human CO2 variations in recent history is wrong and that most of the recent increase in CO2 in the atmosphere indeed is mainly man-made, but that need a step-by-step explanation. Follow the steps:
  1. Accuracy of modern CO2 measurements.
  2. Variations due to local circumstances.
  3. Variations of CO2 due to the seasons.
  4. Where to measure? The concept of "background" CO2 levels.
  5. Evidence of human influence on the increase of CO2 in the atmosphere.
    1. The mass balance
    2. The process characteristics
    3. The 13C/12C ratio
    4. The 14C/12C ratio
    5. The oxygen use
    6. The Ocean's pH and pCO2
  6. Conclusion
  7. Extra: how much human CO2 is in the atmosphere?
  8. References

1. Accuracy of modern CO2 measurements:

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. This is especially the case for the micro-Schollander method, which was intended for exhaling air oxygen and CO2 (around 2%)  measurements.
Charles D. Keeling wanted methods which needed a minimum of maintenance/chemicals for a maximum of near-continuous measurements, which he found in the NDIR (non-dispersive infrared) method.
According to his notes [1][2]:

The imprecision in measuring references gases approaches 0.1 ppmv and is rarely greater than 0.2 ppmv. However, agreement differences less than 0.5 ppmv between flask and analyzers or between different analyzers on a short-term basis are difficult to obtain. Monthly averages from May 1964 to January 1969 may be in error by as much as 1.0 ppmv; but since 1970, systematic error probably does not exceed 0.2 ppmv. The precision of monthly averages is approximately 0.5 ppmv. In summary, monthly and annual averages of the Mauna Loa data are statistically robust and serve as a precise, long-term record of atmospheric CO2 concentrations.

Of course, there is the question of throwing out of "suspect" values when averaging the hourly data. Some data are "suspect", because they differ too much from the (expected) average of the previous (and following) data with more than a given (3 sigma) difference, or when mechanical/physical problems occur, or when there is wind from the wrong direction (land side in the case of coastal stations), etc. In many cases, the source of the deviation is known. All the data still are available in the hourly averages of 4 raw measurements, but are "flagged" and not used for daily, monthly and yearly averages. More on this in next items.

2. Variations due to local circumstances:

There are several local conditions which can interfere with the measurements. In some locations, this is the influence of nearby (local/regional) vegetation and/or combustion of fossil fuels. In other cases, volcanic outgassing may interfere. This is the case for Mauna Loa. From the same source [1]:

Possible ambient error sources at Mauna Loa include volcanic, vegetative, and man-made effects (e.g., vehicular traffic, and industry). Daily peaks in measured concentrations occur because of complex wind currents. Downslope winds often transport CO2 from distant volcanic vents causing elevations in measured CO2 concentrations. Upslope winds during afternoon hours are often low in CO2 because of photosynthetic depletion occurring in sugarcane fields and forests. Vehicular traffic problems (since corrected) caused exaggerated elevations in 1971. Despite these sources of error and contamination, considerable effort has been made to alleviate and detect these sources.

The difference in including and excluding "flagged" data in the final result gives some differences in variability around the seasonal trend, this can be seen e.g. in the hourly averages of Mauna Loa, where the raw data are given (in the early days, average from 4 samples within an hour, nowadays the average of 2x20 minutes 10-second samples. For a detailed explanation of the procedures, see: [21]). Hourly averages of samples which have large variations within an hour, large variations from hour to hour, upwind conditions, etc. are "flagged". Flagged data aren't used for daily, monthly or yearly averages. The data can be downloaded from [3] where MLO = Mauna Loa, the explanation of the data and different flags used in [3] can be found in [4]:

Mauna Loa hourly raw
Mauna Loa raw hourly averages 2004.

Mauna Loa 2004 hourly data selected
Mauna Loa selected (without "flagged") hourly averages 2004.

For 2004, 8784 hourly average data should have been sampled, but:
1102 have no data, due to instrumental errors (including several weeks in June).
1085 were flagged, due to upslope diurnal winds (which have lower values), not used in daily, monthly and yearly averages.
655 had large variability within one hour, were flagged, but still are used in the official averages.
866 had large hour-by-hour variability > 0.25 ppmv, were flagged and not used.

As one can see in the trends, despite the exclusion of (in the above second graph) all outliers, the difference in trend with or without flagged data is minimal, only the number of outliers around the seasonal trend is reduced and the overall increase in 2004 in both cases is about 1.5 ppmv.

That local production/uptake of CO2 has an influence can be seen in the detailed trend of Mauna Loa: with upslope winds, air comes from the valleys where agriculture and other vegetation reduces CO2 levels (with about 4 ppmv) during some parts of the day:

Mauna Loa hourly upslope winds
Mauna Loa hourly averages during 1.5 weeks

This shows that with upslope winds, the data is influenced by local CO2 depleted air. These data are rightfully discarded from the daily/monthly/yearly averages, as they don't reflect the background CO2 levels, which we are interested in.

Does discarding of "contaminated" data influence the trend over a year or several years? I have asked that question to Pieter Tans, responsible for dataprocessing of the Mauna Loa data. His answer:

The data selection method has been described in Thoning et al., J. Geophys. Research, (1989) vol. 94, 8549-8565.  Different data selection methods are compared in that paper, including no selection.  The methods give annual means differing by a few tenths of 1 ppm.  I assume that you have read the README file [4] when downloading the data.  The hourly means are NOT pre-processed, but they are flagged when the of the minute averages is large.

The good, the bad and the ugly stations.
Several stations are deemed "good", as these have minimal influence from local vegetation and/or human emissions (traffic, heating). These are stations in the middle of the oceans, sometimes at coastal points (as long as the wind is not blowing from land side) and/or above the inversion layer. These stations, after discarding outliers, differ from each other within 5 ppmv for yearly averages, of which most is from the delay between the NH and the SH, see next item. 10 of them, spanning the globe from near the North Pole (Alert, Canada) to the South Pole, are used as reference for daily, monthly and yearly averages and yearly trends. The graphs and the data can be found at [5].

Some inland stations, like Schauinsland only give reliable "background" CO2 levels, when the station (at 1200 m altitude) is above the inversion layer and with enough wind speed. This happens only about 10% of the time.

And last, but not least, many inland stations are practically unsuitable for background CO2 measurement, because of incomplete mixing with the higher air layers, partly due to too many local sources/sinks like vegetation and/or human use of fossil fuels, partly due to a shielded location. This is the case for e.g. Diekirch (Luxemburg) [6], where the station is in a valley with forests, urbanisation and traffic in the main upwind direction:

Diekirch (Luxemburg) CO2 measurements compared to wind speed
Graph from [6].

As can be seen, even at inconvenient places with lots of local sources/sinks, 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. The same is true for diurnal variance: at daytime and with high enough wind speed (> 1 m/s), CO2 levels are lower and near background, while at night under the inversion layer, CO2 levels are up to 100 ppmv higher.

3. Variations of CO2 due to the seasons:

There are two main natural influences on the CO2 levels of the atmosphere: the temperature of the ocean's surface waters and the uptake of CO2 by plants in spring/summer and the release of CO2 by the decay of dead plant material in fall/winter. This is most clear in the NH (Northern Hemisphere), where most of vegetation on land is situated.
CO2 is continuously emitted by deep sea upwelling, especially in the tropics, where temperatures are high and the partial pressure of CO2 (pCO2) in the upper oceans is higher than in the atmosphere above it. CO2 is continuously absorbed in the upper ocean layers at higher latitudes, where the colder temperatures reduce the pCO2 of the oceans, lower than the pCO2 of the atmosphere. This is especially the case at the sink places of the THC (thermohaline circulation) in the Nordic Atlantic ocean. Colder water can retain more CO2 than warmer water, but in the case of CO2 there are also a lot of chemical and biological reactions which influence the solubility of CO2 and hence pCO2 at the surface of the oceans. For more details on this, Wiki has a quite good explanation.
The CO2 flow between the tropics and the colder places in the oceans is relatively constant (more about that later), and doesn't influence the seasonal variation that much. More variation is in the temperature (and thus pCO2) of the mid-latitudes, where there is absorption of CO2 in winter and release of CO2 in summer. The CO2 flow of vegetation (including algues in the upper oceans) is in opposite direction: more release in winter and more uptake in summer. The net effect in the NH is a variation of +/-4 ppmv in Mauna Loa (mid Pacific Ocean, middle troposphere) between summer and winter, up to +/-20 ppmv for Barrow (Alaska, USA, sea level, near tundra) or even 35 ppmv at Schauinsland (Germany, 1200 m high). The data of Schauinsland are heavily contaminated by the nearby fully inhabited and industrialised Rhine valley. And influenced by vegetation, in this case the Black Forest of SW Germany. Only at night, when separated from the valleys by an inversion layer, and with sufficient wind speed, the CO2 levels are better mixed with the rest of the troposphere and retained. This is the case for only 10% of the data.
Data series from the SH (Southern Hemisphere) show much less seasonal variation, because of the much smaller area of land/vegetation. The smallest influence of the seasons is found at the South Pole.

Here follows some comparison of the Mauna Loa (selected) monthly averages with these of other stations:

monthly averages 2002-2004
Monthly trends 2002-2004 of 2 NH stations (Barrow and Mauna Loa) and 2 SH stations (Samoa and South Pole)

As can be noticed, the variation at Mauna Loa is smaller than at Barrow and the SH stations have a much smaller seasonal influence than the NH stations. Also, although the trend of the SH stations is near the same, there is some lag between the NH and SH stations. This is the first indication that the source of the increase is situated in the NH, as the ITCZ (intertropical convergence zone) forms a barrier for the exchange of CO2 (and aerosols) between the NH and the SZ. This is even more clear in the longer term yearly trends:

CO2 trends
Trends in yearly averages of CO2 levels at different stations.

The trend of the SH stations has a growing delay of 6-12 months behind the NH trend. But all yearly average data of the "best" stations (and the average of least contaminated data from less suitable stations like Schauinsland) are within 5 ppmv for similar growth.

4. Where to measure? The concept of "background" CO2 levels.

The concept was launched by C.D. Keeling in the mid fifties, when he made several series of measurements in the USA. He found widely varying CO2 levels, sometimes in samples taken as short as 15 minutes from each other. He also noticed that values in widely different places, far away from each other, but taken in the afternoon, were much lower and much closer resembling each other. He thought that this was because in the afternoon, there was more turbulence and the production of CO2 by decaying vegetation and/or emissions was more readily mixed with the overlying air. Fortunately, from the first series on [2], he also measured 13C/12C ratios of the same samples, which did prove that the diurnal variation was from vegetation decay at night, while during the day photosynthesis at one side and turbulence at the other side increased the 13C/12C ratio back to maximum values.

Keeling's first series of samples, taken at Big Sur State Park, showing the diurnal CO2 and d13C cycle was published in [7], original data (of other series too) can be found in [8]:

diurnal variation of CO2 and d13C at Big Sur State Park
Figure 3.1 Diurnal variation in the concentration and carbon isotopic ratio of atmospheric
CO2 in a coastal redwood forest of California, 18-19 May 1955, Big Sur St. Pk.
(Keeling, 1958. Reproduced by permission of Pergamon Press.)

Several others measured CO2 levels/d13C ratios of the their own samples too. This happened at several places in Germany (Heidelberg, Schauinsland, Nord Rhine Westphalia). This confirmed that local production was the origin of the high CO2 levels. The smallest CO2/d13C variations were found in mountain ranges, deserts and near the oceans. The largest in forests, urban neighbourhoods and non-urban, but heavely industrialised neighbourhoods. When the reciproke of CO2 levels were plotted against d13C ratios, this showed a clear relationship between the two. Again from [7]:
CO2 level / d13C correlation
Figure 3.5 Relation between carbon isotope ratio and concentration of atmospheric CO2 in
different air types from measurements summarized in Table 3.4
(Keeling, 1958, 1961: full squares; Esser, 1975: open circles; Freyer and Wiesberg, 1975,
Freyer, 1978c: open squares). All []13C measurements have not been corrected
for N2O contamination (Craig and Keeling, 1963), which is at the most in the area of + 0.6‰

The search for background places.
Keeling then sought for places on earth not (or not much) influenced by local production/uptake, thus far from forests, agriculture and/or urbanisation. He had the opportunity to launch two continuous measurements: at Mauna Loa and at the South Pole. Later, other basic stations were added, all together 10 from near the North Pole (Alert, NWT, Canada) to the South Pole, most of them working continuous, some working with regular flask sampling.

We are interested in CO2 levels in a certain year all over the globe and the trends of the CO2 levels over the years. So, here we are at the definition of the "background" level:
Yearly average data taken from places minimal influenced by vegetation and human sources are deemed "background".
For convenience, the yearly average data from Mauna Loa are used as reference. One could use any base station as reference or the average of the stations, but as all base stations (and a lot of other stations, even Schauinsland) are within 5 ppmv of Mauna Loa, with near identical trends, and that station has the longest near-continuous CO2 record, Mauna Loa is the reference.

As the oceans represent about 70% of the earth's surface, and all oceanic stations show near the same yearly averages and trends, already 70% of the atmosphere shows background behaviour (within 5 ppmv). This can be extended to near the total earth for the part above the inversion layer.

Measurements above the inversion layer.
Above land, diurnal variations are only seen up to 150 m (according to [7]).
Seasonal changes reduce with altitude. This is based on years of flights (1963-1979) in Scandinavia [7] and between Scandinavia and California [9]. Further confirmed by old and modern [10] flights in the USA and Australia (Tasmania). In the SH, the seasonal variation is much smaller and there is a high-altitude to lower altitude gradient, where the high altitude is 1 ppmv richer in CO2 than the lower altitude. This may be caused by the supply of extra CO2 from the NH via the southern branch of the Hadley cell to the upper troposphere in the SH.

From [7], based on [9]:
seasonal CO2 vs. altitude
Figure 3:2 Amplitude and phase shift of seasonal variations in atmospheric CO2
at different altitudes, calculated from direct observations by harmonic analysis
(Bolin and Bischof, 1970. Reproduced by permission of the Swedish Geophysical Society.)

From [10]:
CO2 above inversion layer
Modern flight measurements in Colorado, CO2 levels below the inversion layer
 in forested valleys and above the inversion layer at different altitudes

As one can see, again the values above the inversion layer are near straight and agree within a few ppmv with the Mauna Loa data. Below the inversion layer, the morning values are 15-35 ppmv higher, in the afternoon, these may sink to background again.

If we take the 1000 m as the average upper level for the influence of local disturbances, that represents about 10% of the atmospheric mass. Thus the "background" level can be found at 70% of the earth's air mass (oceans) + 90% of the remaining land surface (27%). That is in 97% of the global air mass. Only 3% of the global air mass contains not-well mixed amounts of CO2, which is only over land.

General conclusion:
Background CO2 levels can be found over all oceans and over land at 1000 m and higher altitudes (in high mountain ranges, this may be higher).

5. Evidence of human influence on the increase of CO2 in the atmosphere.

5.1. The mass balance
The amount of CO2 emitted by humans nowadays is about 7 GtC/yr (CO2 counted as carbon). The increase in the atmosphere is about 4 GtC/yr. That implies that there is little to no increase in the atmosphere due to other causes, or the amount in the atmosphere in the case of a natural unbalance should be higher than the emissions, not lower. To show this for the past near 50 years [11]:

CO2 mass balance
The graph shows the increase in CO2 from emissions, land use and others
 vs. the increase found in the atmosphere, land and oceans, expressed in GtC/yr

The graph is based on calculations of emissions, sampled from national inventories of fuel use and land use change. In the best case, these are accurate, in the worst case, the emissions are underestimated (as probably is the case for China). Inventories of the atmosphere are based on very accurate measurements of CO2 at Mauna Loa. The difference between CO2 emissions (expressed in gigaton carbon per year - GtC/yr) and carbon increase in the atmosphere is what the oceans and/or vegetation absorb each year. The partitioning between land and ocean as sinks is not accurately known, but not of interest here, as in every year the sum of land+oceans is more sink than source.

Some investigations were done to know the sink partitioning between land and oceans by Battle ea. [12], based on changes of d13C and oxygen content in the atmosphere over the last decade of the previous century and earlier estimates:

Battle ea. partitioning land/ocean sink

The increase of CO2 in the atmosphere of about 60 ppmv (122 GtC)  in the past near 50 years is about 60% of the increase since the start of the industrial revolution. This is based on accurate measurements at Mauna Loa and a lot of other places. The amount of CO2 in the atmosphere in pre-industrial times is based on ice cores, which of course are less certain and more smoothed, but there are other proxies with a better resolution in time, which point to lower CO2 levels prior to the emissions.

As the first graph shows, in any year of the past near 50 years, the emissions are larger than the increase in the atmosphere. That means that the total mass balance of all natural variables (temperature, ocean pH, vegetation) which influence CO2 levels, is always towards more sink than source over any year:

Csources + Cemissions = Csinks + dCair
where Cemissions = 2.5 (1960) to 7 GtC/yr (2006) and increasing
and dCair = 1 (1960) to 6 GtC/yr (2006) and increasing in ratio with 55% of the emissions
and Csinks = Csources + 1.5 (1960) to 4 GtC (2006) +/- 2 GtC (natural variability), where the sink capacity increases in ratio with about 45% of the emissions, while the natural variability around the trend seems not increasing or decreasing. The natural variability in sink capacity seems mainly temperature dependent with a ratio of about 4 ppmv/°C, based on temperature/CO2 changes during the 1992 Pinatubo eruption and the 1998 strong El Niño. This is short-term variability around the trend. Over (very) long periods of time (Vostok ice core), the influence of temperature on CO2 levels is about 8 ppmv/°C.

The natural seasonal exchange between vegetation and oceans at one side and the atmosphere at the other side is estimated at about 150 GtC/yr. But that is not of interest for what the change is over a year, as most of the natural releases are absorbed within the same year. The difference after a year is not more than +/- 2 GtC, mainly caused by temperature changes (El Niño, Pinatubo eruption). Thus the natural variations over a year are smaller than the emissions. No matter how high the natural seasonal turnover might be, in all years over the previous near 50 years, the natural CO2 sinks were larger than the natural CO2 sources... Thus it is impossible that natural sources were responsible for (a substantial part of) the increase of CO2 in the past 50 years.

This proves beyond doubt that human emissions are the main cause of the increase of CO2, at least over the past near 50 years. But there are much more indications for that...

5.2. The process characteristics:

Here is a graph of the (global) temperature trend, the cumulative amount of emissions and the increase of CO2 in the atmosphere (1900-2004):

emissions, CO2, temperature increase

The temperature trend is for the sea surface, according to the Hadley Centre in the UK. The emissions are from the international inventory data base. And the CO2 in the atmosphere pre-1959 are from ice cores (Law Dome, Siple Dome), from 1959 on, the data are taken at Mauna Loa (Hawai). Baseline of the CO2 level is 300 ppmv. The 21 year moving average is added, as some have found a good correlation between that average and the CO2 increase. That is right after 1980 (where the correlation was based on), but fails for the whole 1900-2004 period.

As one can calculate, the correlation between temperature and CO2 levels in the atmosphere is rather weak (corr.:  0.843; R^2: 0.711), and from the detailed T/dCO2 graphs, one can see that a huge change in temperature in a certain year has little influence on CO2 levels, compared to the influence of the temperature change over the whole trend.:

correlation CO2 temperature 1900-2004

This indicates that temperature is not the cause of the trend, but the cause of the variation (+/- 1.2 ppmv) around the trend (currently over 1.5 ppmv/yr). More about that is coming (if I had only 36 hours in a day)...

At the other side, the correlation between cumulative emissions and increase in the atmosphere is a near-fit (corr.: 0.998; R^2: 0.997) over the whole period:

correlation emissions - CO2 levels

The ratio is about 0.53% between increase in the atmosphere and what is emitted. This points to a simple linear first order process, directly related to the partial pressure difference between CO2 in the atmosphere and CO2 in the oceans (and vegetation). The higher the CO2 content in the atmosphere, the higher the push to drive CO2 into the oceans. Currenty, the partial pressure difference (pCO2) between the atmosphere and the oeans is about 7 ppmv [18], based on ship's surveys and buoys. The increase of the pCO2 difference causes more and more uptake of CO2 by the oceans (and similarly in vegetation).

Again, this clear relationship points to a direct influence of the emissions on the increase in the atmosphere. There is no known natural process that is able to force CO2 into the atmosphere exactly in the same ratio as is the case here for the emissions. One only need to look at the difference in variability of the temperature curve (which has a limited influence on CO2 levels of about 4 ppmv/°C) with the smoothness of the emissions curve...

This adds to the weight of the emissions as main cause of the increase in the atmosphere.

5.3. The 13C/12C ratio:

The carbon of CO2 is composed of different isotopes. Most is of the lighter type: 12C (that has 6 protons and 6 neutrons in the kernel), and about 1.1% is heavier: 13C (has 6 protons and 7 neutrons in the kernel). There also is some 14C (6 protons and 8 neutrons in the kernel), which is continuously formed in the upper stratosphere from the collisions of nitrogen with cosmic rays particles. This type of carbon (also formed by above-ground atomic bomb experiments in the 1950's) is radio-active and can be used to determine the age of fossils up to about 60,000 years.

One can measure the 13C/12C ratio and compare it to a standard. The standard was some type of carbonate rock, called Pee Dee Belemnite (PDB). When the standard rock was exhausted, this was replaced by a zero definition in a Vienna conference, therefore the new standard is called the VPDB (Vienna PDB). Every carbon containing part of any subject can be measured for its 13C/12C ratio. The comparison with the standard is expressed as d13C in per mille:

(13C/12C)sampled – (13C/12C)standard
——————————––––––––––––––– x 1.000

Where the standard is defined as 0.0112372 part of 13C to 1 part of total carbon. Thus positive values have more 13C, negative values have less 13C. Now, the interesting point is that vegetation growth in general uses by preference 12C, thus if you measure d13C in vegetation, you will see that it has quite low d13C values. As fossil fuels were formed from vegetation (or methanogenic bacteria, with similar preferences), these have low d13C values too.  Most other carbon sources (oceans, carbonate rock wearing, volcanic outgassing,...) have higher d13C values. For a nice introduction of the isotope cycle in nature, see the web page of Anton Uriarte Cantolla [13].

This is an interesting feature, as we can determine if CO2 levels in the atmosphere (which is at -8 per mille VPDB) were influenced by vegetation decay or fossil fuel burning (both about -24 per mille) at one side (towards the negative side) or by ocean degassing (0 to +4 per mille) towards the positive side as largest possible sources.

From different CO2 base stations, we not only have CO2 measurements, but also d13C measurements. Although only over a period of about 25 years, the trend is clear and indicates an extra source of low d13C in the atmosphere.

d13C trends
Recent trends in d13C from direct measurements of ambient air at 10 base stations. Data from [14].
ALT=Alert; BAR=Barrow; LJO=La Jolla; MLO=Mauna Loa; CUM=Cape Kumukahi; CHR=Christmas Island;
SAM=Samoa; KER=Kermadec Island; NZD=New Zealand (Baring Head); SPO=South Pole.

Again, we see a lag in the trends with altitude and NH/SH border transfer and less variability in the SH. Again, this points to a source in the NH. If that is from vegetation decay (more present in the NH than in the SH) and/or from fossil fuel burning (90% in the NH) is solved in the investigation of Battle ea. [12], where it is shown that there is less oxygen used than can be calculated from fossil fuel burning. Vegetation thus produces O2, by incorporating more CO2 than is formed by decaying vegetation (which uses oxygen). This means that more 12C is incorporated, and thus more 13C is left behind in the atmosphere. Vegetation is thus a source of 13C and is not the cause of decreasing d13C ratio's.

And we have several other, older measurements of d13C in the atmosphere: ice cores and firn (not completely closed air bubbles in the snow/ice). These align smoothless with the recent air measurements. There is a similar line of measurements from coralline sponges and sediments in the upper oceans. Coralline sponges grow in shallow waters and their sceleton is built from CO2 in the upper ocean waters, without altering the 13C/12C ratio in seawater at the time of building. The combination of atmospheric/firn/ice and ocean measurements gives a nice history of d13C changes over the past 600 years:
coralline sponges d13C
Figure from [15] gives a comparison of upper ocean water and atmospheric d13C changes.

What we can see, is that the d13C levels as well as in the atmosphere as in the upper oceans start to decrease from 1850 on, that is at the start of the industrial revolution. In the 400 years before, there is only a small variation, probably caused by the temperature drop in the Little Ice Age.

Again this is a good indication of the influence of fossil fuel burning...

5.4. The 14C/12C ratio:

14C is a carbon isotope that is made in the atmosphere by the impact of cosmic rays. It is an unstable (radioactive) isotope and breaks down with a half-life time of about 6,000 years. 14C is used for radiocarbon dating of not too old fossils (maximum 60,000 years). The amount of 14C in the atmosphere is variable (depends of the sun's activity), but despite that, it allowes to have a reasonable dating method. Until humans started to burn fossil fuels...

The amounts of 14C in the atmosphere and in vegetation is more or less in equilibrium (as is the case for 13C: a slight depletion, due to 12C preference of the biological reactions). But about halve of it returns to the atmosphere within a year, by decay of leaves. Other parts need more time, but a lot is going back into the atmosphere within a few decades. For the oceans, the lag between 14C going into the oceans (at the North Atlantic sink place of the great conveyor belt) is 500-1500 years, which gives a slight depletion of 14C, together with some very old carbonate going into solution which is completely 14C depleted. In pre-industrial times, there was an equilibrium between cosmogenic 14C production and oceanic depletion.

Fossil fuels at the moment of formation (either wood for coal or plankton for oil) incorporated some 14C, but as these are millions of years old, there is no 14C anymore left. Just as is the case for 13C, the amount of CO2 released from fossil fuel burning diluted the 14C content of the atmosphere. This caused problems for carbon dating from about 1890 on. Therefore a correction table is used to correct samples of after 1890. In the 1950's another human intervention caused trouble for carbon dating: nuclear bomb testing induced a lot of radiation, which nearly doubled the atmospheric 14C content. Since then, the amount is fast reducing, as the oceans replace it with "normal" 14C levels. The half life time is about 5 years.

Again, this adds to the evidence that fossil fuel burning is the main cause of the increase of CO2 in the atmosphere...

5.5 The oxygen use:

To burn fossil fuels, you need oxygen. As for every type of fuel the ratio of oxygen use to fuel use is known, it is possible to calculate the total amount of oxygen which is used by fossil fuel burning. At the other hand, the real amount of oxygen which is used can be measured in the atmosphere. This is quite a challenging problem, as the change in  atmospheric O2 from year to year is quite low, compared to the total amount of O2 (a few ppmv in over 200,000 ppmv). Moreover, as good as for CO2 as for oxygen, there is the seasonal to year-by-year influence of vegetation growth and decay. Only since the 1990's, oxygen measurements with sufficient resolution are available. These revealed that there was less oxygen used than was calculated from fossil fuel use. This points to vegetation growth as source of extra O2, thus vegetation is a sink of CO2, at least since 1990. The combination of O2 and d13C measurements allowed Battle e.a. [12] to calculate how much CO2 was absorbed by vegetation and how much by the oceans (see 5.1). The trends of O2 and CO2 in the period 1990-2000 can be plotted in a nice diagram:

bolin graph
O2-CO2 trends 1990-2000, figure from the IPCC TAR [16]

This doesn't directly prove that all the CO2 increase in the atmosphere is from fossil fuel burning, but as both the oceans and vegetation show a net uptake, and other sources are much slower and/or smaller (rock weathering, volcanic outgassing,...):

There is only one fast possible source: fossil fuel burning.

5.6 The ocean's pH and pCO2:

If CO2 is increasing in the atmosphere with about 55% of the accumulated emissions,a part is absorbed by vegetation (see 5.4), another part is absorbed by the oceans. When CO2 is absorbed by the oceans, this is partially in solution in its original form, but some of it reacts with available carbonate ions to form bicarbonate. Between 1751 and 1994 the average surface ocean pH is estimated to have decreased from approximately 8.179 to 8.104 (a change of -0.075), based on recent and older oceanic surveys [17].

But the ocean's pH can be interpreted the other way out: if for whatever reason the pH is reduced (e.g. by an undersea volcanic event with lots of SO2), this leads to an important increase of pCO2(oceans) and may release a lot of oceanic CO2 into the atmosphere. But that is in contradiction with the observed change in 13C: if oceanic CO2 (from the deep oceans to the surface and further into the atmosphere) was released, this should increase the 13C content of both the upper oceans and the atmosphere, while we see the reverse. Moreover, the release of more CO2 from the upper oceans due to a lower pH would reduce the total amount of DIC in the water. But we see the reverse trend: DIC is increasing over time [22]. Thus the increase of atmospheric CO2 is going into the oceans, not the reverse.

Further, another part of the oceanic survey compares the pCO2 of the atmosphere with that of the oceans at different latitudes in different oceans. This shows that there are huge changes in oceanic pCO2 at different latitudes due to changes in temperature and DIC (dissolved inorganic carbon, that is CO2 + bicarbonate + carbonate). This gives a permanent release of CO2 in the tropics (pCO2 of 750 µatm in the upper oceans vs. 385 µatm for the atmosphere) and a permanent sink of CO2 in the polar oceans, especially in the North Atlantic (150 µatm vs. 385 µatm). The oceans at mid-latitudes are seasonal emitters/absorbers of CO2, depending of the water temperature and sea life (plankton). The average difference of pCO2(atm) - pCO2(oceans) is about 7 ppmv. That means that in average more CO2 is going from the atmosphere into the oceans than reverse [18]. Moreover, different surveys over time revealed that ocean parts which were net sources of CO2 gradually changed into net absorbers [19].

Although the ocean pCO2 data are scattered in time and covered area the trends are clear that the average (increasing) flow of CO2 is from the atmosphere into the oceans and not the reverse.

This adds to the overall evidence that human emissions are the main cause of the increase of CO2 in the atmosphere.

6. Conclusion

From the available evidence it is quite clear that human emissions are the main cause of the increase of CO2 in the atmosphere. There is a small influence of temperature on this increase, as warmer oceans emit some CO2 (but warmer land absorbs more CO2 in vegetation!). But the influence of temperature is limited: based on the variability of the CO2 increase around the trend, the short-term (1-6 months) ratio is about 3 ppmv/ºC (based on the 1992 Pinatubo and 1998 El Niño events). The very long term influence of temperature on CO2 levels (Vostok ice core) is about 8 ppmv/ºC. Thus at maximum, the influence of temperature on the current increase is 0.7 ºC x 8 ppmv/ºC = 5.6 ppmv of the about 100 ppmv increase since the start of the industrial revolution.

There are only two fast main sources of CO2 to the atmosphere, besides the burning of fossil fuels: oceans and vegetation. Vegetation is not a source of CO2, as the oxygen deficiency (in 5.5) showed. Neither are the oceans, as the 13C trend (in 5.3) and the pCO2/pH (in 5.6) shows. This is more than sufficient to be sure that human emissions are the cause of most of the increase of CO2 in the atmosphere of the past 1.5 century.

Thus we may conclude:

All observed evidence from measurements all over the earth show with overwhelming evidence that humans are causing the bulk of the increase of CO2 into the atmosphere.


That humans are the cause of the recent increase of CO2 doesn't tell anything about the influence of increased CO2 on temperature!

7. Extra: how much human CO2 is in the atmosphere?

A lot of people is confused about this point: Only a few percent of the atmosphere is currently from human origin. That is because every year about 150 GtC of CO2 (somewhat less than 20% of the CO2 content)  is exchanged between the atmosphere and the oceans/vegetation. That means that every single CO2 molecule from human origin has a 20% chance per year to be incorporated in vegetation or dissolved into the oceans. This makes that the half life time of human CO2 in the atmosphere is only about 5 years. This was confirmed by the fate of 14C, increased due to atomic bomb testing, after the tests stopped. Thus if humans emit 8 GtC in a given year, next year some 6.5 GtC is still of human origin, the rest was exchanged with CO2 from the oceans and vegetation. The second year, this still is 5.3 GtC, then 4.3 GtC, etc... This is not completely accurate, as some of the "human" CO2 comes back next year(s), especially from vegetation, as much of vegetation is one-year old leaves, which rotting returns a high part of CO2 incorporated in previous years. This is less the case for the oceans, where more of the absorbed CO2 disappears into the deep oceans, where it isn't directly traceable anymore. There are techniques to follow human CO2 even there, where they use other recent human-made gases like CFC's to track the past emissions. Anyway the "half life", that is the time period in which half of the human induced CO2 disappears, is about 5.2 years.

Over longer periods, humans continue to emit (currently about 8 GtC) CO2. The accumulation over the last years thus is 8 + 5.3 + 4.3 + 3.5 + 2.8 +... or about 40 GtC from the emissions over the past 30 years. That is only 5% of the current atmosphere...
Some conclude from this that humans are only responsible for 5% of the CO2 increase and thus, as far as that influences temperature, also only for 5% of the temperature increase. But that is a wrong assumption...

The previous paragraphs are about how much human induced CO2 still is in the atmosphere. That is about the origin and fate of individual CO2 molecules, which atmospheric lifetime is governed by the seasonal turnover (back and forth flows) of about 150 GtC in/out the atmosphere from/to oceans and vegetation, and has nothing to do with the fate of the extra amount of CO2 (as mass) that humans emit, neither with the increase of total amount of CO2 in the atmosphere as result of that. The latter is governed by the net amounts which year by year are incorporated into oceans and vegetation. That is only 1-7 GtC/year (variable due to temperature variability) or in average about 55% of the emissions. The half life time of this extra CO2 (as mass) is much longer than the half life time of an individual CO2 molecule: around 40 years [20]. Thus if we should stop all CO2 emissions today, then the increase of 100 ppmv since the start of the industrial revolution would be reduced to 50 ppmv after some 40 years, further to 25 ppmv after 80 years and 12.5 ppmv after 120 years...

The IPCC comes with much longer half life times, according to the Bern model. This is a combination of relative fast (upper oceans), slower (deep oceans) and very slow (rock weathering) sinks for the extra CO2. They assume that the first, relative fast, sink of CO2 will reduce in capacity over the years. Some media talk about hundreds to thousands of years that the extra CO2 will be in the atmosphere. That is true for the last part of the curve, as the smaller amounts of CO2 are getting slower and slower into the sinks. But the bulk (87.5 %) of the extra CO2 will disappear within 120 years.

From several discussions, I know that it is quite difficult to understand the two different mechanisms which govern the fate of human CO2 in the atmosphere: the fate of individual molecules, governed by exchange rates (turnover) and the fate of an increase in total CO2, governed by absorption rates (sink capacity). Here I try to give an example of how to interprete the difference:

Let us say that you start the day in your shop with € 1000.00 in your cash register, all 1000 euro is in 1 euro pieces, all stamped in France. During the day, you have about € 200.00 expenses from goods delivery and you receive € 192.00 back from sales. At the end of the day, you have € 992.00 in your cash register, not only with French euro's anymore, but part of them are now stamped in Germany, Belgium, Spain,...
Next day, you add some € 16.00 from your own personal money, only euro's stamped in The Netherlands, to the cash register to start a fresh day with € 1008.00. During that day the same happens as in the previous day: € 200.00 expenses, € 192.00 income. Thus the day ends with € 1000.00 in your cash register, with now an increase of Netherlands euro's (but less than what you have added). Next day, you add € 16.00, again in Netherlands euro's and end the day with € 1008.00. You can repeat that for a few weeks, until you run out of personal money... Over several weeks, you will see that the number of euro's from The Netherlands slowly increases in ratio, but that the increase of the total amount in the cash register is only 50% of what you add on a daily base. That means that you have a problem: your expenses are larger than your income. That also means that despite the huge daily exchanges (which result in a rapid reduction of Netherlands euro's), that has no influence at all on the total amount of money you have at the end of the day, only what you have added yourself and the (negative) difference of the total balance counts. In this case there is no (net) addition of money from your daily bussiness, only a daily loss.

The difference between the two half life times of CO2 is comparable to at one side the fate of the number of Netherlands euro's in the cash register at the end of each day (which depends of the amounts which were added and exchanged that day and the composition of the exchanges), while on the other side, the second half time only depends of the total sum of euro's that is added and what rests from all tarnsactions at the end of the day. That is independent of the height of each individual transaction or the number of transactions, or the composition of the transactions: the total loss/gain at the end of the day is what you have earned or lost that day... In this case, there is a continuous loss of CO2 (as quantity!) added by humans, which means that all natural flows of CO2 in/out the atmosphere together, over a full year, gives zero net addition to the atmosphere: nature acts as a sink for human CO2...

As shown in chapter 5, there is little doubt that humans are fully responsible for most of the increase of CO2 in the past (at least halve) century, that means that - as far as there is an influence of CO2 on temperature - that humans may be responsible for (a part of) the temperature increase. How much, that is an entirely different question, as that mainly depends of the (positive and negative) feedbacks that follows any increase of temperature...

8. References

[1] Carbon Dioxide Concentrations at Mauna Loa Observatory, Hawaii, 1958-1986, CDIAC NDP-001:

[2] Rewards and penalties of monitoring the earth, Charles D. Keeling, Ann. Rev. Energy. Envir. 1998.23.25-82:
Fascinating autobiographic story from C.D.Keeling about the history of CO2 measurements and the struggle against the administrations to get and continue funding for continuous measurements.
Of special interest:
- First measurements on 5 l flasks were done with enhanced barometric equipment, with an accuracy of better than 0.1 ppmv.
- The same barometric equipment was used to test calibration gases and NIR equipment. A change in calibration gases (air/CO2 vs. N2/CO2) caused a jump in response of the NIR equipment. All previous collected data were corrected for this change.

[3] Hourly average data of continuous (2 x 20 minutes 10 second snapshots + 3 calibrations per hour) CO2 measurements at different stations:

[4] Explanation of the different data and flags used in the records of [3]:

[5] Selected monthly average data and graphs of CO2 measurements of the 10 base stations:

[6] Seasonal and Diurnal CO2 Patterns at Diekirch, Luxemburg:

[7] Scope 13 - The global CO2 cycle, H.-D. Freyer, SCOPE Workshop on the Biogeochemical Cycling of Carbon held at Ratzeburg, Federal Republic of Germany, 21-26 March 1977
Although the data series are outdated (much more extended and much more is known nowadays), the principles outlined and the original findings still are of interest.

[8] The concentration and isotopic abundance of carbon dioxyde in rural areas, C.D. Keeling, 1957. Can be found at the web site of Ernst Beck ( ).
His earlier work at Big Sur is here:

[9] The CO2 Content of the Upper Polar Troposphere between 1963-1979, W. Bischof, (1980?):
Data and trends are within 2 ppmv of Mauna Loa. Seasonal variability around the trend is +/- 4 ppmv

[10] The Airborne Carbon in the Mountains Experiment (ACME): Initial Modeling Results, Stephan F.J. De Wekker ea. Powerpoint presentation (made around 2004). (1.45 MB .ppt file).

[11] State of knowledge on the vulnerability of the oceanic CO2 sink, Powerpoint presentation by Corinne Le Quéré, British Antarctic Survey and University of East Anglia. (2.52 MB .ppt file).

[12] Global Carbon Sinks and Their Variability Inferred from Atmospheric O2 and d13C, Battle ea., Science, Vol. 287 31 March 2000.
More up-to-date (Bender e.a.):

[13] The 13C cycle in nature,
Anton Uriarte Cantolla. Some of the figures and text are from the book "Historia del Clima de la Tierra" (Spanish).

Atmospheric Carbon Dioxide and Carbon Isotope Records at CDIAC:

Evidence for preindustrial variations in the marine surface water carbonate system from coralline sponges, Böhm ea., GEOCHEMISTRY GEOPHYSICS GEOSYSTEMS, 2002.

[16] The Carbon Cycle and Atmospheric Carbon Dioxide, IPCC TAR, Chapter 3,  page 206, fig. 3.4
[17] The acidification of the oceans, according to Wikipedia
Based on investigations done by Key e.a. and Orr e.a.

[18] Uptake and Storage of Carbon Dioxide in the Ocean: The Global CO2 Survey, Feely e.a. on the NOAA pages:

[19] See the (large: 10 MB) powerpoint presentation by Metzl at the pCO2 workshop (2007):

[20] Calculations by Peter Dietze at the web site of John Daly:

[21] Detailed procedures of measurements, calibrations and calculations for the Mauna Loa (and other baseline station) CO2 data:

Interannual to Decadal Variability of the Ocean Carbon Cycle  from the Bermuda Institute of Ocean Sciences

On the net: 16 September, 2007.
Last update: 15 February, 2010.

To the
family home page

To the climate change page

Zend naar: