NOAA CO2 data from 2021 are not yet finalized as I write this.  According to NOAA for 2020, net CO2 from all sources and sinks, human and natural, on average increased only 2.58 ppm for the year 2020.  That is only 0.000258% of the atmosphere and it includes all CO2 from all sources and sinks, natural and human. 

In other words, human-produced CO2 emitted to the atmosphere minus human-produced CO2 absorbed by the environment DID NOT EXCEED a 0.000258% increase for 2020.  Net global CO2 atmospheric concentration measured at Mauna Loa (MLO) was 414.24 ppm. (That’s the annual mean for MLO in 2020 minus annual mean for MLO in 2019.  414.24 ppm minus 411.66 ppm equals 2.58 ppm.)

Then 2.58 ppm divided by 414.24 ppm is 0.0062 or 0.6%.  This means that the annual increase for 2020 in total CO2 is only 0.6% of the total CO2 in air and this 0.6% percent increase includes CO2 from all sources and sinks, natural and human.  In other words, the net increase of CO2 for 2020 due to humans did not exceed 0.6% of the total CO2 in air.

NOAA, etc. writes statements such as:

“From 2000 through 2018, CO2 emissions to the atmosphere from burning of fossil fuels rose from 6.7 PgC yr ^–1 to 10.2 PgC yr^-1 (1 petagram of carbon is 10¹⁵ gC, or 1 billion metric tons C, or 3.67 billion metric tons CO2). Global fossil fuel emissions have increased steadily year upon year, with the exception of 2009 following the global economic recession and 2014-2016 when emissions held nearly constant (Figure 1).”  https://gml.noaa.gov/ccgg/carbontracker/index.php#north_america

NOAA and global warming proponents typically neglect to inform the reader that CO2 added to the air in excess of the Henry’s Law partition ratio is absorbed by the environment.  They also neglect to inform that the MLO-measured-and-reported CO2 concentration, the defacto “gold standard”, is in fact the net residual difference between two very much larger natural CO2 fluxes, CO2 emission flux and CO2 absorption flux.  The apparent annual increase (i.e., the MLO rate of change of net global CO2 atmospheric concentration, aka the Keeling Curve slope) cannot be caused by humans, as shown below.  Near the bottom of the same page linked above, NOAA informs the reader of important uncertainties in their estimate of CO2 emission flux, including that it contains guesswork.  This may be the most useful information on that website. 

There are many natural sources of CO2 and many natural sinks for CO2 and both are orders of magnitude larger than human CO2 emissions.  In the real world, the net global average CO2 concentration in air (and even more so the net human CO2 emission) cannot be distinguished from noise in the measurement system.  In the real world away from computer models, atmospheric CO2 concentration is determined by nature; humans can neither increase it or decrease net global average CO2 concentration except as a temporary perturbation.

CO2 measurements typically ignore the variability of both CO2 and air in the natural environment.  Mauna Loa (MLO) is essentially a lab environment; its measurements are diligently made, accurate and precise for that lab location.  The unit they use is ppm, that is CO2 parts per million parts of air.  As measured by NOAA, it is micromoles of CO2 gas per mole of dried air, which is the same as ppm.  But it (ppm) is a ratio of masses, not a volume measurement. 

Thus, there are at least two major problems in the “gold standard” MLO measurement of net global average CO2 atmospheric concentration as commonly, and repeatedly reported worldwide.  First, they have removed a huge variability in CO2 data by removing water and water vapor from the sample.  CO2 gas is highly absorbed by water, the concentration of water and water vapor in air is more than 10 times larger than CO2, and humidity is highly variable.  In practice, if they do not remove the water from the air samples, then the variability in the data is so large that it prevents the CO2 measurement; for this reason they use a molar mass measurement (ppm) instead of a volume measurement such as micrograms CO2 per liter of air (or ppmv); ppm and ppmv are not equivalent units.

Secondly, estimations of the mass of the atmosphere vary tremendously.  The denominator in ppm is mass of the atmosphere or portions thereof.  This mass is highly variable, but the uncertainty implied by that variability is almost never propagated to the ppm ratio.  This second problem is a partial derivative of the first problem above, but also there are additional variables.

Global warming proponents Kevin E. Trenberth and Christian J. Guillemot (1994) state in The total mass of the atmosphere: https://agupubs.onlinelibrary.wiley.com/doi/abs/10.1029/94JD02043

“Thus the mean mass of water vapor is 1.25×10¹⁶ kg and the dry air mass is 5.132×10¹⁸ kg, corresponding to a mean surface pressure of 982.4 mbar. Overall uncertainties are ∼0.1 mbar or 0.5×10¹⁵ kg in total mass and about double those values for atmospheric moisture content.” 

These global warming proponents point out that the mass variability is high for dry air, but variability of moisture content is about double of that mass variability.

According to this estimate (Trenberth and Guillemot, 1994), the atmosphere contains about 5.132 X 10¹⁸ kilograms of dry air, which is 5.132 X 10²¹ grams of dry air.  Then 3.9 X 10^-6 grams of CO2 (the annual increase per gram of dry air) multiplied times the estimated 5.132 X 10²¹ grams of dry air in the total atmosphere results in an estimated increase in the atmosphere of 2.0 X 10¹⁶ grams of CO2 for 2019-2020 due to all sources and sinks, natural and human.  Sounds like a big, scary increase.  But wait, there is more that is rarely if ever mentioned.

Thus we have 2.0 X 10¹⁶ grams of CO2 added to the atmosphere (from all sources natural and human), which are diluted immediately and continuously into 2 natural, continuous, opposite vector-directional fluxes of CO2.  One flux is CO2 absorb into the environment, primarily ocean surface, which is 71% of earth’s surface.  The other flux is CO2 emission flux, also primarily from ocean surface.  CO2 gas is continuously colliding with earth’s surface, thus continuously being simultaneously emitted and absorbed at all normal earth temperatures.  Although we do not precisely know the size of these two continuous CO2 vector fluxes, the MLO-measurement (i.e., the Keeling Curve) is the record of the rate of change of the net annual difference between these two giant fluxes, i.e., 2.0 X 10¹⁶ grams of CO2 per year for 2019-2020, which is the slope of the Keeling curve expressed in grams of CO2 per year for 2019-2020. 

The above “Keeling Curve” would appear flat (as in the graph further above) if the left hand axis represented the whole atmosphere instead of 0.01% of the atmosphere.

Notice, for the estimated mass of the atmosphere (5.132 X 10²¹ grams of dry air), that the net annual increase in CO2 (i.e., the MLO-measured net annual difference between the two fluxes, or 2.0 X 10¹⁶ grams of CO2 for the year for 2019-2020) is more than 5 times larger than the estimated average annual fossil fuel CO2 emissions (3.67 X 10¹⁵ grams of fossil fuel CO2 emissions for 2000 through 2018, from the NOAA reference above).  This is CO2 fossil fuel emissions only, not net emissions.  Net emission is emissions minus absorptions; net CO2 fossil-fuel emissions would be about half of the estimated 3.67 X 10¹⁵ grams of average annual fossil fuel CO2 emissions.  The two growth curves (2.0 X 10¹⁶ grams of net CO2 per year versus 3.67 X 10¹⁵ grams of fossil fuel CO2 emissions per year) are diverging over time and the net global CO2 is growing faster.  Therefore, it is not scientifically plausible that CO2 emissions from humans burning fossil-fuels are causing the slope of the Keeling curve, i.e. the net increase in global CO2 atmospheric concentration. 

The very slowly increasing slope of (a) the “gold standard” measured net global CO2 atmospheric concentration (i.e. the NOAA-Scripps Institute lab at Mauna Loa) cannot be caused by (b) the more slowly increasing slope of estimated CO2 emissions from fossil fuels combustion because the slope of (b) is less than (a) when (b) and (a) are on the same scale. The slopes are diverging with respect to time. There are no exceptions.

Also notice, for the estimated mass of the atmosphere (5.132 X 10²¹ grams of dry air), that the estimated uncertainty is 0.5 X 10¹⁸ grams, (i.e., 0.5×10¹⁵ kg from the above reference Trenberth and Guillemot, 1994).  The uncertainty in the denominator of ppm (i.e., 0.5 X 10¹⁸ grams) is about 25 times larger than the numerator (2.0 X 10¹⁶ grams) AND this large uncertainty has not been propagated to the resulting ppm ratio.  In other words, the ~414 ppm measurement, although precisely measured in the lab sample, is highly uncertain in the real world; it does not represent the high variability of CO2 concentration in normal atmosphere.  With an uncertainty (or standard deviation) which is 25 times larger than the measured CO2 sample amount, it is highly improbable that the ~400 ppm signal could be distinguished from noise in a sufficiently powered sampling of the natural environment.   Here is a reference on propagation of uncertainty:  https://chem.libretexts.org/Bookshelves/Analytical_Chemistry/Supplemental_Modules_(Analytical_Chemistry)/Quantifying_Nature/Significant_Digits/Propagation_of_Error