A Closer Look at CO2 Growth

For a while I have been looking at atmospheric carbon dioxide data from stations around the world.  This post draws together some observations, many of which are pretty much common knowledge- but some of what I’ve found is surprising.

So I’ll start by listing some of this common and not so common knowledge:-

-The often quoted figures for global CO2 levels are not at all global, but are the local readings at Mauna Loa in Hawaii.

-The long term carbon dioxide record shows continuing increase at all stations, indicating greater output than sinks can absorb. 

-Southern Hemisphere CO2 concentration is increasing but more slowly than the Northern Hemisphere.  Their trends are diverging.

-Seasonal peaks in CO2 concentration occur in late winter and spring in both hemispheres.

-There is very great inter-annual variation in the seasonal cycle of CO2, which can be even more than the average annual increase.

-This inter-annual variation occurs at the same time in both hemispheres, even though the seasonal cycles are 6 months apart.  This implies a global cause, such as the El Nino Southern Oscillation (ENSO).  Large volcanic eruptions also have an impact.  There are likely to be other factors.

-Sea surface temperature change precedes CO2 change by 12 to 24 months.  It is difficult to reconcile this with ocean out-gassing as a cause of the inter-annual CO2 changes.  It is nonsense to claim that CO2 change leads to sea surface temperature change.

-ENSO changes occur at about the same time as CO2 changes.

-CO2 concentration increases during La Ninas. 

-El Ninos precede higher sea temperatures by 4 to 6 months.

-Because of the “oscillation” part of ENSO events, strong events are followed by opposite conditions 16 to 24 months later.  In this way a strong El Nino will lead to strong ocean warming often followed by La Nina conditions and higher CO2 concentration.

-The slowing Southern Hemisphere trend and flattening curve at the South Pole lacks satisfactory explanation.

CO2 measuring stations

Geoffrey Sherrington has shown differences existing between NOAA and Scripps daily CO2 data at Mauna Loa, and that uncertainty in daily data must be much greater than the claimed 0.2 part per million.  His article confirmed my decision to use Scripps instead of NOAA data.  In this post I use Scripps monthly data from many stations across the Pacific, and data from the CSIRO station at Cape Grim in Tasmania, to compare observations from different locations.

Figure 1 shows the locations of stations in the Scripps network, and Cape Grim.

Figure 1:  Scripps stations and Cape Grim

Point Barrow is the most northerly part of the USA, and Alert is the most northerly part of Canada.

The often quoted figures for global CO2 levels are not at all global.  They are not the global average, nor are they representative of other locations.  They are in fact the local CO2 concentration from the slopes of Mauna Loa in Hawaii.  The trend in CO2 increase is similar to, but not the same as, those in other locations.

Figure 2 shows monthly CO2 concentrations from all of the Scripps stations.

Figure 2:  Monthly CO2 at all locations

It is clear that all stations show a similar rising trend, and all show seasonal variation of varying degrees.  However, few stations have long term records, and most have periods of missing data. 

Differences, similarities, and divergence

Figure 3 shows monthly differences from the Mauna Loa record of stations with fairly complete records. 

Figure 3:  Six stations’ difference from Mauna Loa

Monthly differences show huge seasonal variation, so Figure 4 shows 12 month average differences.

 Figure 4:  Six stations’ difference from Mauna Loa, 12 month averages

Clearly, there are major differences between the different records: 

-La Jolla has too many gaps for further analysis. 

-There are differences between Cape Grim and South Pole from about 1980 to the early 1990s.

-Southern Hemisphere stations (American Samoa, Cape Grim, and South Pole) are diverging from Mauna Loa, and from Barrow Point and Alert.  Figure 5 shows these trends more clearly.

Figure 5:  Barrow Point and South Pole difference from Mauna Loa, 12 month averages

While South Pole and Mauna Loa are strongly diverging, Barrow Point and Mauna Loa are becoming slightly more similar.

In Figure 6, the divergence of South Pole data is evident in monthly readings.

Figure 6:  Monthly CO2 concentrations, Mauna Loa, Barrow Point, and South Pole

Note how much larger the Barrow Point seasonal range is.  More importantly, note how South Pole data begin well within the Mauna Loa range, but 50 years later barely reach the bottom of the Mauna Loa range, as Figures 7 and 8 show.

Figure 7:  Monthly CO2 concentrations, Mauna Loa and South Pole 1965-1975

Figure 8:  Monthly CO2 concentrations, Mauna Loa and South Pole 2010 -2020

Why the divergence?  How can a well-mixed gas show a lower trend at the South Pole?  Why is it that the South Pole summer draw down has decreased and is now a plateauing?

Seasonal change

Now zooming in to look at seasonal swings in just two years, 2011 and 2012:

Figure 9:  Monthly CO2 concentrations, Mauna Loa, Barrow Point and South Pole

The Barrow Point range from low to high is nearly three times the size of the Mauna Loa range, and the South Pole range is tiny.  The peak concentrations at Barrow Point and Mauna Loa are in late spring, with a sharp drop at Barrow Point to August and a smoother curve at Mauna Loa to lows in autumn; while at the South Pole the annual curve is better described as a shallow rise in winter followed by a “peak” in spring and a long plateau over summer, with a very small decrease in late summer.  The next three plots show the timing of highs and lows at these three stations for the whole record.

Figure 10:  Timing of seasonal high and low CO2 concentrations, Mauna Loa

Annual lows are in September or October, and highs are almost always in May.

Figure 11:  Timing of seasonal high and low CO2 concentrations, Barrow Point

Lows are always in August, while highs are spread across late winter to late spring, with a plateau from February to May (and extending twice into June).

Figure 12:  Timing of seasonal high and low CO2 concentrations, South Pole

At the South Pole, seasonal highs are reached in spring or early summer, with lows in late summer and early autumn, with one instance in June.

Inter-annual changes

While the seasonal cycles appear to be regular, the timing and size of seasonal changes can vary considerably from year to year.

The next plots show detrended data since 1985 for several locations (few have good data before 1985).  Detrending allows us to compare inter-annual variation more easily.  We do this for each record by subtracting the trend.

Figure 13:  Detrended monthly CO2, Mauna Loa

Figure 14:  Detrended monthly CO2, Barrow Point and Alert

Figure 15:  Detrended monthly CO2, South Pole and Cape Grim

While the seasonal range is different for each location, there is remarkable similarity in timing of changes, for example the late 1980s- early 1990s and 2009-2013.  Note how close Cape Grim and South Pole are, although Cape Grim is at 40.68 degrees South, 49 degrees north of the South Pole.  The South Pole data appear to be representative of a large part of the Southern Ocean.

Because the detrended data retain enormous seasonal variations, it is necessary to show the detrended data (this time from 1979) with monthly means subtracted, for Barrow Point in the far north, Mauna Loa in the middle, and South Pole at the extreme south.  Here are the seasonal signals:

Figure 16: Seasonal signals of monthly CO2 data

As an example, Figure 17 compares detrended data from Barrow Point with monthly means:

Figure 17:  Detrended monthly CO2 with monthly means, Barrow Point

Subtracting the monthly means shows the residual variation in carbon dioxide for Barrow Point:

Figure 18:  Detrended monthly CO2 with seasonal signal removed, Barrow Point

Figure 19 combines the three stations:

All three records follow the same pattern, with a large increase from 1979 to the late 1980s, followed by decrease in the 1990s.  There appears to be another steep increase from 2012 to the present.  Notice that Mauna Loa and South Pole values can be from 1 ppm below to 2 ppm above the trend, while at Barrow Point the range can be from 4ppm below to 5 ppm above the trend, which is about 2.5 ppm per year. 

However, there is still a large amount of variation in the monthly figures.  A centred 13 month rolling mean makes comparison much easier.

Figure 20:  Centred 13 month mean of detrended monthly CO2 with seasonal signal removed

The similar pattern followed by stations from north to south, from the Arctic Ocean, across the Pacific, to the Antarctic, far from any industrial or cropping contamination, is immediately obvious.  The Barrow Point record appears to lag behind Mauna Loa and South Pole data by from one to five months.  South Pole can be a few months ahead to a few months behind Mauna Loa, even though South Pole absolute monthly concentration peaks are from four to seven months later.

Ocean temperature effects

In Figure 14 of my post on 2nd May, Will Covid-19 Affect Carbon Dioxide Levels? I showed that CO2 change lags one year behind sea surface temperatures (SSTs).  The next plot shows the centred 13 month mean of HadSST4 data, scaled up to compare with CO2 data.

Figure 21:  Scaled, centred 13 month mean of detrended monthly HadSST4 and CO2 data with seasonal signal removed

Now the same data with SSTs lagged 12 months…

Figure 22:  Scaled, centred 13 month mean of detrended monthly HadSST4 and CO2 data with seasonal signal removed, HadSST4 lagged 12 months

Large change in CO2 concentrations appears closely linked with sea surface temperature a year before- (or even two years, as between 2002 and 2010).  Sea surface temperatures have a global effect.

ENSO effects

Another cause of CO2 variation is the El Nino- Southern Oscillation (ENSO) which appears in the swings between El Nino and La Nina conditions.  ENSO has a great effect on weather conditions globally, affecting winds, clouds, rainfall and temperature.  Figure 18 shows how CO2 levels respond to the Southern Oscillation Index (SOI), which is a good indicator of ENSO conditions.

Figure 23:  Centred 13 month means, scaled SOI and detrended CO2 levels

CO2 increases in La Ninas.  The pattern becomes more intriguing when we plot inverted SOI levels with sea surface temperatures, as in Figure 19.

Figure 24:  Scaled, centred 13 month mean of detrended monthly HadSST4 with seasonal signal removed and scaled inverted SOI

Inverted SOI data indicate SST data 4 to 6 months later.  (The early 1980s and early 1990s don’t match because of the huge volcanic eruptions of El Chichon and Pinatubo.)  In other words, an El Nino will raise ocean temperatures, and a La Nina will lower ocean temperatures, 6 months later.  Because of the oscillating nature of ENSO, El Ninos and La Ninas approximately reflect each other 16 to 24 months later, as Figure 20 shows.  (Again, El Chichon and Pinatubo have a large impact.)

Figure 25:  Scaled SOI, normal and inverted

That pattern recurs, with varying lag times, throughout the whole 144 year SOI history.

Which is why SSTs will probably increase to about February of 2021…

Figure 26:  Scaled SOI, normal and inverted, and detrended HadSST4

…and with them, CO2 concentration.

Figure 27:  Scaled SOI, normal and inverted, and detrended HadSST4, with South Pole CO2 data

This image has an empty alt attribute; its file name is soi-inv-sst-co2-1.jpg


The long term carbon dioxide record shows continuing increase at all stations, indicating greater output than sinks can absorb. 

CO2 concentrations and trends, while similar, have discernible differences at different locations, notably between the hemispheres.

CO2 concentrations at Southern Hemisphere stations are increasing, but more slowly than those in the Northern Hemisphere, such that their trends are diverging.

On the long term CO2 rise are seasonal rises and falls, most likely due to seasonal vegetation, crop, and phytoplankton growth and decay. 

Peaks in CO2 concentration occur after winter and spring in both hemispheres- February to May at Barrow Point, April and May at Mauna Loa, and September-December at the South Pole.  This is not due however to a six month delay in CO2 mixing from sources in the Northern Hemisphere to the Southern, otherwise the South Pole trend would be the same.  It is lower, and becoming more so. 

There is great variety in seasonal range of CO2 at different locations, with greatest variation in the Arctic and the least in the Southern Hemisphere.

The amount and timing of these seasonal rises and falls varies from year to year.  These inter-year changes in CO2 concentrations can be as much as or greater than the normal annual increase.

Even though the South Pole station is far from the Southern Ocean, especially in winter when sea ice extends further, and even further from any vegetated land areas, its data appear representative of a great part of the Southern Hemisphere.

Small inter-annual changes in sea surface temperatures have a large impact on these changes in CO2 concentrations at South Pole and Mauna Loa about 12 to 24 months later.  There can be a further delay of up to five months in the effect at Point Barrow. 

This is not controversial.  According to the CSIRO, these variations “have been shown to correlate significantly with the regular El Niño-Southern Oscillation (ENSO) phenomenon and with major volcanic eruptions. These variations in carbon dioxide are small compared to the regular annual cycle, but can make a difference to the observed year-by-year increase in carbon dioxide.”

While sea surface temperature rise precedes CO2 concentration increase, there is no evidence at all of CO2 concentration change preceding sea surface temperature change.

With an apparent approximate 12 – 24 month delay between ocean temperature change and inter-annual CO2 change, changes in ocean out-gassing and absorption rates appears to be an unlikely mechanism.  Changes in land vegetation, forests, crops, and oceanic phytoplankton, moderated by the changing circulation, rainfall, cloud, and temperature patterns of ENSO events, appears to be a more likely mechanism, with the much smaller land area of the Southern Hemisphere accounting for the much smaller changes. 

The unresolved problem

This does not however explain the decreasing amount of summer draw down at the South Pole, and the divergence from Northern Hemisphere data.   Perhaps Southern Ocean phytoplankton are not decreasing as much during winter, so the CO2 sink is slightly increasing, slowing the CO2 growth trend a little and smoothing the CO2 growth curve.  Who knows?  I have yet to see a satisfactory- or any- explanation.

Tags: , , , , , , ,

7 Responses to “A Closer Look at CO2 Growth”

  1. Bill In Oz Says:

    Well Ken that is a big mouthful… But a couple of key facts do emerge..That CO2 increases and decreases as the Lan Nina / El Nino cycle proceeds.

    I suggest that the IOD cycle also has a huge impact on CO2 as well..
    And whatever similar process happens in the Atlantic oceans..

    So for me the key question re CO2 concentrations is dependent on working out WHY those massive processes happen in the planet’s oceans.

    Such processes release and are powered by a huge amount of energy. Where does that huge amount of energy come from ? Solar insolation does not change that much from year to year so I doubt it is the source.

    And a conclusion from all this seems clear : we humans don’t have much impact on the levels of CO2 in the planet’s atmosphere.

  2. ozdad Says:

    Thank you for your detailed analysis again .

  3. Big M Says:

    Your last two sentences really summarise science. “Who knows?”

  4. TonyN Says:


    I was given the link to your impressive paper by Geoff Sherrington.

    I think I have the answer to the mystery of the CO2 annual variations, which is simply that they have got their sums wrong from the very beginning. I apologise for the very elementary step-by-step style of writing, but as the error is so egregious yet has been misssed by everyone involved, as with paradoxes and illusions of all kinds including magic card tricks, they take a step-by-step deconstruction to reveal the ‘trick’

    As described in the following links, NOAA/Scripps run a continuous process at various sites which takes samples of the local atmosphere and changes the proportions of the constituent gases by removing one of the gases, namely water vapour, and measures the quantity of CO2 molecules in the remaining dried sample.

    Conversion of this number into a dimensionless ratio of mole fraction in ‘parts per million by volume’ requires division by the number of all of the other molecules in the dried sample.
    NOAA/Scripps don’t do this but use the denominator number from their prepared dry gases used in the calibration process.



    Whilst the resulting percentage of CO2 holds for dry air, it cannot be applied as a percentage of CO2 in the real atmosphere, which contains a highly variable quantity of water vapour whose proportion is affected by temperature and pressure, and so can range from near-zero to over 30%. They subtract this gas,thus reducing the volume of the sampe but effectively increasing the percentage of all other gases in proportion to the amount of water-vapour that they extracted from the original sample, before measuring the CO2 content!

    If as I believe the NOAA /Scripps result for dry air is being used as a proxy for the CO2 content of the real atmosphere, the conjecture is made that the NOAA/Scripps process overstates the CO2 ratio, in proportion to the humidity of the original air sample.

    This conjecture can be tested by looking at the NOAA’s Point Barrow (PTB) CO2 concentration trend graph, which shows an annual variation in CO2 of circa 20ppm, or ~ 5%


    Whilst the NOAA data for this site does not include humidity, pressure, or temperature from which the annual change in water vapour content can be established, the average annual monthly meteorological data data is available from the nearby airport 15km away:


    This website reports the hottest and coldest annual average monthly climate data for Point Barrow [PTB] as :

    Jul avg T 5 degC RH 87% Barometric Pressure 1014mb *Abs Humidity 7.4%

    Jan avg T -25 degC RH 79% Barometric Pressure 1020mb *Abs Humidity 0.49%

    * Conversion from Relative Humidity to Absolute Humidity ( % of water vapour by volume) is made using the tool on this website:

    The result indicates that as the average water vapour content rises from a winter low to a summer high, the CO2 also rises by roughly the same amount.

    This is is a remarkable result but can’t easily be cross-checked against the other sites due to the difficult of finding the monthly average meteorological data. But we can look at the CO2 result from Samoa (SAM) in the southern tropics, where although the annual variation in CO2 reading is only 1 ppm or so, we can compare it with the average highest and lowest annual monthly atmospheric water vapour amounts at nearby Pago Pago airport:

    Jul avg T 27degC RH 81% Barometric Pressure 1012mb *Abs Humidity 30.6%

    Jan avg T 28 degC RH 82% Barometric Pressure 1008mb *Abs Humidity 30.7%

    If the conjecture were true, one would then expect to see a variation of 0.1%, or 0.4 ppm, which is very close to the reported ~1 ppm annual variation.

    What the conjecture also predicts is that as the absolute humidity at SAM is fairly constant over the year at circa 30.6%, the CO2 measurements at SAM will be overstated by a whopping 30%!

    If further work shows that this conjecture were to be accepted as true, the consequences would be:

    The ‘Keeling Curve’ annual variation in CO2 is an artefact, which varies with variation in humidity, and hence seasonal temperature, and is not caused by N. Hemisphere vegetation growth and decay as postulated by Keeling, NOAA and accepted by many others.

    The Scripps/NOAA process contains a hidden bias and the output data can be in error by up 30% and more

    – Any climate modelling based on NOAA/Scripps data will be invalidated by this huge hidden variable error. This bias if integrated and subtracted from the IPCC’s total world CO2 concentration, the whole Zero Carbon movement would lose credibility.

    Reported rising CO2 content is actually caused by rising temperature, and not the other way round. Therefore, the NOAA/Scripps process is effectively measuring atmospheric temperature changes.

    – The effect of CO2 on changing the climate is far less than currently believed.

    To summarise, we must look elsewhere for the major cause of rising temperatures.

    Well, Ken (and others) I would appreciate a critique. I may be utterly wrong, but that is citizen science for you.


    Tony Nordberg (aka TonyN on WUWT)

    • kenskingdom Says:

      Wow!! and Hmmm!! My first thoughts: But at the south pole, while the annual seasonal variation is minimal, the long term trend of CO2 is not much less than for Mauna Loa, and Antarctica is definitely not warming very much at all?? And we should also see CO2 levels reflecting temperatures e.g The Pause and not trailing land and sea temperature changes by many months??

      • Bill In Oz Says:

        Ken as discussed above 1) the Mauna Loa site dat ii=s suspect, Who with any professional competence, ever puts CO2 measuring station on the slope of bloody big active volcano ? 2) Yes Antarctica is not warming but the CO2 measured there will reflect what is released elsewhere on the planet and then well mixed by the global wind systems.

      • tony Says:


        Thanks for the reply, it has made my day!

        Whilst my conjecture points to the overstatement of total CO2 content by a potentially large amount, and the consequent effect on the sensitivity ratio could show that Zero Carbon policies are boojums and not snarks , there remains the apparent long-term rise as you say.

        It seems to be going up at ~0.5% per year or so, and I wonder about the existence of a systematic tiny bias that, if integrated in the data processing of huge number of readings over the year, could cause such a rise.

        I have been sceptical of the use of commercial IR gas analysers, as over 60 years this technology has changed a lot. Firstly, they have to freeze-dry the sample, as these analysers can’t cope with moisture, which consequently they make a virtue out of necessity by talking about  ‘ppmv in dry air’ rather than ppmv in real air.

        As I understand it in the early days there were problems with the IR analyser as muck built up on the glass window, causing a dimming … and as we know the system is based on IR absorbtion so such a dimming will cause a progressive over-reading with time; as would ageing of the light-source if incandescent lamps were used in the early instruments. Then, there is the replacement of the whole instrument which must have happened at least once, which could cause  a discontinuity in the record requiring a smoothing. The same kind of discontinuity could have occurred when they changed the material of the calibration pressure flasks, I speculate for reasons of diffusion, or maybe corrosion, as for example a reduction in oxygen would cause a corresponding over-reading of CO2 content.

        Together, it is likely that there have been a few retrospective ‘adjustments’ of the previous record, to preserve the story.

        Now as we are looking at apparent long-term rise of 1.5ppm  per year, we can start with a change of IR detector reading over the year of some 1.5 millivolts. This error of ~1.5% of full-scale of +5v or so over that time could easily be accounted for in terms of noise , thermal drift , ageing of components, and so on.

        Anyway, for the next step they they take the non-linear quantum, and ‘correct’ it to make it linear, and here we have more potential sources of error.

        But my money is on an accumulation of quite small rounding errors at the Digital-to-Analog conversion stage as well as in the computing stage, which will accumulate through totalising. For example, at 1 reading per hour we are looking at ~8760 readings per year. To report a running total we need to add the current reading and then divide by the number of current readings. Effectively we are integrating the delta, and as the number of counts goes up through the year, any rounding errors will also accumulate. If we are looking for roughly a 1.5ppm per year long-term increase, we would only need a positive rounding error of 1.7 in 10,000, per reading.  Given a 5v reference, this would equate to 0.5 millivolts, or the least-significant-bit of a perfect 14-Bit ADC!

        This is all obviously speculative, and only full access to the whole system and it’s evolution over the past 60 years, including the manuals of all of the analysers etc, plus local environmental build-up (e.g. South Pole communities) , all of the versions of the software, plus the information on the computers used (remember the Intel 286 arithmetical error disaster?). God knows how much the analysis would cost, but maybe given the world cost of decarbonising, it would be worth engaging a University with say five or so cheap postgrad labour, if only to justify the expense. However, who would want a PHD in demolishing lucrative beliefs?


Comments are closed.

%d bloggers like this: