I started this exploration into the changes in atmospheric CO2 because I was surprised that a widely reported significant reduction in anthropogenic emissions in 2020, resulting from the COVID-19 pandemic curtailing industrial activity, commuting, and air travel, did not result in a discernable decline in the rate of growth of atmospheric CO2. I was not satisfied with the official explanations for the impact not being observed. I decided to examine the data and see if I could tease out the effects, or understand better why not, with more-detailed graphs.
While several sources estimate that the decline in anthropogenic CO2 emissions averaged about 7% to 10% for all of 2020, Carbon Brief has an interactive map that shows the global reduction reached over 18% in mid-April 2020; it was down to about 10% in March and May. Coincidentally, those months are near the usual, annual peak of CO2 Winter ramp-up, when Northern Hemisphere photosynthesis is at a minimum. One would naively expect that, if anthropogenic CO2 were important in the growth of the atmospheric concentration, double-digit percentage drops in the three last months of the annual 8-month ramp-up would at least be suggested in the height of the peak (range) or the slope of the curve. Indeed, it is common to observe ‘benches’ on the growth curves (see Figures 1 and 2, below), related to decreased CO2 flux, perhaps resulting from abnormally low Winter temperatures. However, the ‘bench’ present in 2020 is almost indistinguishable from 2019, and the Winter ramp-up range ties for 3rd place (Presented with one significant figure to the right of the decimal point.) with three other years over the 31-year period analyzed. The rate decline in 2020 was in March, not April, and the April value was a little higher (0.3 PPMv) than in 2019. It is reasonable that there may be a slight lag in observed effects; however, it is X-Files spooky to suggest that a hiatus precedes the decrease in the forcing agent. Therefore, it is impossible that the decline in anthropogenic CO2 flux is responsible for the March decline, which happens frequently.
According to a NOAA Research News Article, “The global rate of increase [2020 CO2] was the fifth-highest in NOAA’s 63-year record, following 1987, 1998, 2015 and 2016.” The same article shows a time-series graph for methane, which shows a seasonal variation also, albeit of different shape. The article makes the following statement with regard to methane:
“Methane in the atmosphere is generated by many different sources, such as fossil fuel development and use, decay of organic matter in wetlands, and as a byproduct of livestock farming. Determining which specific sources are responsible for variations in methane annual increase is difficult. Preliminary analysis of carbon isotopic composition of methane in the NOAA air samples done by the Institute of Arctic and Alpine Research at the University of Colorado, indicates that it is likely that a primary driver of the increased methane burden comes from biological sources of methane such as wetlands or livestock rather than thermogenic sources like oil and gas production and use.”
Considering the obvious seasonality of methane and CO2, I’m inclined to believe that it is temperature, hence weather and climate, that is driving the changes.
I used NOAA data from the Mauna Loa CO2 observatory for analysis. Mauna Loa data are generally considered to be the best representative sample of what is happening in the Northern Hemisphere (NH).
Fig. 1. Recent, typical data from Mauna Loa Observatory (https://gml.noaa.gov/ccgg/trends/) The red lines and symbols represent the monthly mean values, centered on the middle of each month. The black lines and symbols represent the same, after correction for the average seasonal cycle.
Note the saw tooth-like form (red line) resulting from seasonal variations. If a positive trend-line is added to a sine wave (a function of the form y = m x + a sin x), the slope of the increasing side of the sine wave will increase while the absolute value of the slope of the decreasing side of the sine wave will decrease as m increases. That is the opposite of what is observed with the CO2 concentrations! The absolute value of the drawdown slope is greater than the ramp-up slope! By removing the ‘seasonality’ in the black line in Figure 1, it makes it look like there is a trend, presumably driven by continuous fossil fuel emissions, that is modulated by natural variations. I’m making the case that the apparent trend is the result of natural seasonal variations and that the accepted model is not physical.
What I missed, initially, is that the saw tooth pattern is not symmetrical. The ramp-up phase lasts longer than the drawdown phase, although the slope is about two-thirds. In addition, the drawdown phase does not return the CO2 concentration to the original starting level!
A fundamental question to be answered is whether the long-term changes in CO2 concentration are 1) caused by increasing anthropogenic CO2, and by implication, responsible for warming; or 2) if the warming is responsible for rising CO2 levels; or 3) the apparent correlation is spurious and it is just coincidence that both variables are increasing over time.
I created yearly graphs for the years 1990 through 2021, for analysis, by normalizing the beginning of each yearly ramp-up phase by subtracting the October CO2 value from all the CO2 concentrations for the contiguous 12-month period (Oct.–Sept.). Thus, the ramp-up phase starts at zero in all graphs, and all subsequent concentrations are relative to a zero baseline. See Figure 2, below. Some might be concerned that I defined the ramp-up phase as beginning with the lowest, and ending with the highest seasonal values. I did this to emphasize the part of the year when the natural sinks were minimal and there was the greatest hope for detecting a decline in anthropogenic CO2.
Fig. 2. L) Ramp-up and R) drawdown phases for the period of October 2019 through September 2020.
The first half of the yearly cycle is a change from a net absorption of CO2 as plants become senescent and shut down photosynthesis. The annual minimum usually occurs in October; however, rarely, the September average is slightly lower. After the Fall minimum, the CO2 concentration climbs throughout the Winter and early-Spring. The peak is almost always in May; however, it is sometimes as early as April. There are some variations in the pattern in that there is usually a short hiatus for a month or two during the Winter, varying from December to March, as shown in Figure 2, above. It appears that slightly warmer than usual weather may also result in a temporary convex-upward bulge, as shown in Figure 2 above, and conversely, slightly cooler weather results in a concave-upward depression, or bench.
The ramp-up phase is followed by an abrupt decline in the CO2 levels with Spring phytoplankton blooms and leafing-out of deciduous trees and other plants. While biogenic decomposition of dead organic material, producing CO2, is always occurring, the photosynthetic activity on land and in the oceans results in a net decline during the Northern Hemisphere growing season. In addition, theoretically, the reduction in polar pack ice should withdraw additional CO2;however, it isn’t evident.
The 31-year aggregate ramp-up range varies from a low of 6.2 PPMv (in 1999–2000) to a high of 9.4 (in 2015–2016, El Niño); the average is 7.7, with a standard deviation of 0.7 PPMv. The plot of the yearly trend-slopes [not shown] of the ramp-up strongly resembles the range-trend slope (Figure 5) for the 31-year interval. The minimum slope is 0.92 PPMv per month (1999–2000), to a maximum slope of 1.3 PPMv per month (in 2015–2016, El Niño), with an average of 1.1 PPMv per month and standard deviation of 0.1 PPMv.
The shapes of the drawdown curves are much simpler and consistent than the ramp-up phase curves. In addition, the average coefficient of determination (R2) for a linear fit is slightly higher than for the ramp-up curves.
The 30-year [2021 is not available] annual range and slope values for the drawdown phase have R2 values low enough (<0.04) to raise a question as to whether the long-term trend in range and slope (<0.009 PPMv/yr) is statistically significant. The calculated p-values for range and slope are both >0.3; using a p-value cut-off of 0.05, we cannot reject the null-hypothesis that there is no formal predictive value in the independent variable, time. Therefore, the trend is not statistically significant, and we cannot say that the annual drawdown is increasing or decreasing. This suggests that the rate of drawdown is insensitive to the temperature changes. However, we can still say that based on past history, it is probable that the annual drawdown values will continue to oscillate around the current average values (Range: -5.2 PPMv ±0.8; Slope: -1.8 PPMv per month ±0.3) with a 2σ uncertainty.
The wrinkles in the faces of older men are sometimes said to add character. One might similarly say that the CO2 ramp-up curves express more character than the drawdown curves. I think that the transient deviations from a linear growth tells us that something other than the usually constant anthropogenic emissions are having an obvious impact on the yearly growth. Figure 3, below, compares the last 6 years of the ramp-up phase. The El Niño event stands out; however, the other years – except 2017–2018 – are very similar.
Fig. 3. Stacked CO2 ramp-up phase curves.
One might argue that the main driver of the upward trend of CO2 is either increased anthropogenic emissions, which is the default claim, or one could argue that the active sinks simply aren’t keeping up with changes in the temperature-driven emissions. The uncompensated residual at the end of the Summer drawdown is on average about 2.5 PPMv, for the 30-year period. Thus, the next season starts at a base-line that is elevated above the previous season base-line. The drawdown phase is not compensating for the increasing ramp-up. Figure 4, following, illustrates this.
Fig. 4. Annual uncompensated, or residual CO2 at the end of Summer drawdown.
A clue that temperature is driving the annual variations, is that the infamous El Niño events of 1997–1998 and 2015–2016, present as prominent peaks in the graph of the seasonal ramp-up range-trend, as shown below in Figure 5. Note that the 1997–1998 event is also unprecedented in Figure 4, and followed by a similarly unprecedented decline!
Fig. 5. Trend of annual ramp-up phase ranges with peaks at years 7 and 25,
corresponding to the El Niño years of 1997—1998 and 2015–2016, respectively.
If CO2 concentrations were driving temperatures, then one would expect the temperature growth to remain high after a spike in the annual CO2 ramp-up phase. Instead, one observes a sharp decline in temperatures in the years immediately after an El Niño, along with a decline in CO2 concentrations! Clearly, temperature is driving the ramp-up phase variations.
The system seeks an equilibrium between all the sources and sinks, as determined by the solubility of CO2 in water, which is determined by the temperature and atmospheric partial pressure. Bacteria and fungi will increase their activity with increasing temperatures, at temperatures above freezing, also.
Different international agencies have proclaimed 2020 to be either tied with 2016 as the warmest year on record, or the second warmest year. Interestingly, the ‘bench’ is muted in 2020, when the pandemic closures were most severe! This suggests to me that Northern Hemisphere temperatures are more important than anthropogenic emissions.
To explore this further, I obtained GISTEMP v4 Northern Hemisphere-mean monthly, seasonal, and annual means for the Northern Hemisphere, land and ocean. I then calculated the average temperature anomaly for the CO2 ramp-up phases (Oct.–May) for the years 1990 through 2020. The following graph, Figure 6, is the result of plotting the annual range of the ramp-up phase against the temperature anomalies:
Fig. 6. Annual ramp-up phase range versus temperature-anomalies for the years 1990 through 2020.
The R2 value indicates that nearly 44% of the variance in the annual range can be explained by average seasonal temperature changes. This provides better prediction than the time-series, which only has an R2 value of about 28%. I suspect that if multi-variate analysis were to be applied, with the ocean and land temperatures treated as separate variables instead of being averaged, a better fit would result and provide more insight on whether outgassing or biological decomposition were the major contributor. One might want to include an index for upwelling, but I think that would be very difficult to obtain.
When the relationship between CO2 and temperature is examined at a finer temporal resolution than annually, i.e. at monthly averages for a single ramp-up phase (i.e. 2018–2019), the R2 value is reduced significantly to 0.230. This could be explained by differences in lag times between outgassing and bacterial production. However, the graph shows two outliers of large jumps [not shown] in CO2 that are suggestive of transient events that are not temperature dependent. What comes to mind is wildfires, but they rarely occur in the Winter. Also, the transient forcing can’t be anthropogenic because large sudden changes are exceedingly rare, and when they do occur, they aren’t measurable!
There are claims that Spring is coming earlier, and there is evidence that Winters are becoming warmer. Is the current behavior of the ramp-up phase representative of the longer-term behavior? To answer that question, I also looked at three years of early atmospheric CO2 data, 1959, 1960, and 1961, also obtained from the NOAA interactive graph cited below. The maximum and minimum seasonal values occur in the same months as currently, and the shape of the curves are similar to the recent curves, and the slopes and ranges, while lower than most recent data, are within the observed variations for the last 30 years. Thus, I would conclude that there have been no significant changes in roughly the last 60 years. Instead, we are dealing with small long-term changes, not at all unlike the phenomena of NASA ‘mission creep.’ Compare Figure 7, below, with Figure 2, above.
Fig. 7. L) Ramp-up and R) drawdown phases for the period of October 1959 through September 1960.
The slope of a line is important because it is a direct expression of the rate of change of the dependent variable, in this case, CO2 concentration. The slope of the Ordinary Least-Squares Linear Regression trend line for Winter 2019–2020 (1.17 PPMv/month) was higher than the 30-year average (1.09) ramp-up, and higher than the ramp-up slope in 2018–2019 (1.15). Not only can a reduction in range or slope of the growth curve, resulting from a reduction in anthropogenic CO2 not be observed, but it is actually suggesting the opposite!
It is generally claimed that about half-of the anthropogenic CO2 goes into the atmosphere and is totally responsible for the annual increase of about 2 PPMv annually.
Therefore, that would predict a decline in the slope of the CO2 concentrations for the 2019–2020 ramp-up phase of about 9% in April (It increased!), or a lowered April 2020 concentration of at least 0.1 PPMv because of reduced CO2 emissions. It increased compared to 2019 and 2021!
If it weren’t for the economic importance of fossil fuels, we wouldn’t have an estimate of their annual production, consumption, and resultant emissions. The available atmospheric CO2 measurements wouldn’t allow us to make such estimates. That is, not only is the anthropogenic release swamped by natural sources, but even subtle changes, such as a decrease in the rate of increase during the seasonal ramp-up phase, cannot be discerned. The working hypothesis of climatologists is that the long-term atmospheric CO2 increase is the result of anthropogenic emissions. However, the evidence supporting that is weak.
In the finest tradition of post-modern climatology, I will speculate, with little supporting evidence, that the benches in the ramp-up curves during the coldest months of Winter (commonly starting in February, as seen above in Figures 1, 2, 3, and 7.), may suppress oceanic out-gassing, and/or inhibit the activity of biogenic decomposition of detritus, thus causing a hiatus in the increase of CO2. I will further speculate that upwelling plays a role in the temporary increases and decreases in the ramp-up phase.
If the Winter ramp-up is driven by soil respiration, biogenic oxidative decomposition of surface detritus, and ocean out-gassing, it is difficult to imagine an equilibrium between the ramp-up and drawdown when the ramp-up lasts twice as long as the Summer drawdown. An exception might be made if bacterial/fungal action were to be shut down during most of the ramp-up phase by temperatures well below freezing, as might be expected during glaciation. Thus, we could be observing a delayed rebound from the Little Ice Age.
However, because the total anthropogenic contribution to the atmosphere is only about 4% of the total carbon flux into the atmosphere, humans can’t be responsible for this yearly imbalance! The atmosphere can’t tell the difference between anthropogenic sources and natural sources, such as out-gassing and biological decomposition. It is just coincidence that the long-term rise is about one-half of the anthropogenic contributions to the atmosphere.
If there was a delicate balance between the carbon fluxes going into and out of the atmosphere, hydrosphere, and biosphere, then it is conceivable that a perturbation created by Man might cause a reaction to dampen that perturbation, with the effect being to suppress all the anthropogenic-induced change. However, this analysis suggests that because of the annual variations in the ramp-up phase, that delicate balance doesn’t exist, and probably hasn’t existed during at least the last 60 years. The available empirical observations do not support the idea that changes in anthropogenic fluxes are measurable directly. Instead, it appears that temperature is the controlling factor.
However, it does appear that there may be a negative-feedback mechanism, which generally is not recognized. As the Arctic pack ice melts in the Summer, it exposes the underlying cold water to the atmosphere. This allows CO2 to dissolve into the water, reducing the local concentration, and contributing to the general NH drawdown. It is noteworthy that the Fall-Winter CO2 ramp-up phase coincides with the growth of the Arctic pack ice.
It is difficult for me to accept that there is an unrestrained, positive feedback loop driven by CO2 and resulting in significant surface temperature increase, because, if that were the case, one would expect that we would have long ago passed the so-called ‘Tipping Point’ and be in a permanent ‘hot house’ state, like Venus.
It appears that the long-term growth in the atmospheric CO2 concentration is driven not by anthropogenic emissions, but instead, by static effectiveness of the sinks, which because of the seasonal effects, appears to not be keeping up with increasing temperature-driven emissions. There is no question that anthropogenic CO2 is being absorbed in the atmosphere. However, there is no obvious evidence to support the claim that it is totally responsible for the annual CO2 increases. I’m speculating that the carbon flux is large enough that, in the absence of anthropogenic CO2, the annual increase would be at least 96% of what is being measured. The temperature-driven transients are undeniable, and therefore the annual temperature increases must be primarily responsible for the annual increases.
For additional background on this, I can recommend this article by Chaamjamal.
I used NOAA CO2 data from the last 31 years to produce Excel graphs to explore the annual range and slope from the ramp-up period from October through May; additionally, I prepared graphs for the years 1959, 1960, and 1961. The downloaded ASCII data that I used for analysis only covered the period 1973 through 2019. I had to retrieve the last three years, and early-1960s data, manually from the following interactive graph: https://gml.noaa.gov/ccgg/trends/graph.html
The temperature-anomaly data were from the following:
GISTEMP Team, 2021: GISS Surface Temperature Analysis (GISTEMP), version 4. NASA Goddard Institute for Space Studies. Dataset accessed 2021-06-04 at https://data.giss.nasa.gov/gistemp/
Lenssen, N., G. Schmidt, J. Hansen, M. Menne, A. Persin, R. Ruedy, and D. Zyss, 2019: Improvements in the GISTEMP uncertainty model. J. Geophys. Res. Atmos., 124, no. 12, 6307-6326, doi:10.1029/2018JD029522.