You are on page 1of 58

Is the Recent Increase in Atmospheric Carbon Dioxide Natural?

An Attribution Study through Simple but Critical Data Analysis

Jonathan J. Drake

First edition: 10/3/2013

Is the Recent Increase in Atmospheric Carbon Dioxide Natural? An Attribution Study through Simple but Critical Data Analysis

Most of the analysis in this document was done in mid-2009 but lack of time has prevented this work being completed to a standard considered suitable for publication. Some sections, primarily appendices are previously published material. Rather than let this work sit on the shelf, it is offered here as a rough draft. It may evolve as/if time permits.


1. Abstract 2. Introduction 3. Nomenclature 4. Basic Correlation 5. Accounting for Temperature 6. Correlation of Detail 7. Ratios 8. Temperature Lead or Lag? 9. Periodicity in Correlation between Temperature and CO2 10. Accounting for Temperature with Greater Accuracy 11. Cumulation with Half-life & Absorption Parameters 12. Isotopic Changes 13C/12C 13. Conclusions 14. Data & References 15. Appendix 1: Examples of Simple Cumulation on Time Series 16. Appendix 2: Changes to MLO CO2 Record 17. Appendix 3: Additional UAH Temperatures & dCO2/dt relationships 18. Appendix 4: Compilation of CO2 Atmospheric Residence Time 19. Appendix 5: Bern Carbon Cycle Model

Is the Recent Increase in Atmospheric Carbon Dioxide Natural? An Attribution Study through Simple but Critical Data Analysis
Jonathan J. Drake


Assessments of emissions, carbon dioxide and temperature data are made in an attempt to determine the cause of the so-called “anthropogenic rise” in CO2. Several different simple approaches are utilised to search for possible scenarios and a potential attribution.

It is shown that models based upon accumulation of emissions can readily produce spurious results and therefore tight constraints must be placed upon any associated controlling parameters, ideally from independent measurements. Impulse responses and carbon dioxide residence times are examined in the appendices. The long residence times leading to the assertions that emissions accumulate in the atmosphere are questioned and appear to be based upon instantaneous doubling which is unrealistic. Other related topics are also touched upon.

A relatively high degree of correspondence appears to be present between North Atlantic Oscillation (NAO) and CO2 levels suggesting that the primary cause of increased atmospheric carbon dioxide is more likely to be natural processes, principally ocean out-gassing, rather than emissions from the burning of geologically derived hydrocarbons.

Prevailing orthodox assumptions about humans being the primary cause of contemporary increases in atmospheric carbon dioxide do not appear to be supported by the data and should be reassessed.


There have been many analyses of carbon dioxide levels in relation to man-made carbon dioxide emissions over the years. Many take a cursory look at the trends and assert they are related, simply because both have upward trends. Others go a step further and create atmospheric carbon dioxide time series from emissions data. Most use mathematical models and methods that seem well informed and justified. However, it has been noted that many of the processes employed can, by inappropriate selection of parameters, produce high correlations to the atmospheric carbon dioxide trend from almost any dataset containing positive and increasing values. This is particularly true of techniques using a cumulative construction with partial absorption and half-life parameters.

Throughout this study, it will be assumed that the atmospheric levels of CO 2 can be attributed to a combination of natural and human related effects.

It is the noted that hydrocarbon burning emissions are estimates and may have significant uncertainties. However, this does not stop climate scientists and economists using the data in their models and therefore it should suffice for this study. Also, due to politics, financial interests and other factors, it is conceivable that emissions figures are actually underestimated. This will not be discussed further but should have little if any impact upon the presented arguments and conclusions.

This study will start by looking at whether there is any causal relationship between emissions and rise in atmospheric CO2 and then explore other associated topics.


        

Concentration of CO2 = CO2 Annual change in CO2 = dCO2/dt Rate of change of dCO2/dt = d2CO2/dt2 Emissions = Annual emissions from hydrocarbon burning Emissions are expressed in metric tons (tonnes) of carbon, unless otherwise noted 1 million metric tons (Mt) of carbon Mt of carbon is written as Mt C and less formally as Mt 1000 million metric tons of carbon (Gt C) Annual change in emissions = dEmissions/dt (Mt/year/year)

Basic Correlation

Firstly an assessment of global carbon dioxide (CO2) and emissions will be made, as depicted in Figure 1. It is obvious by looking at the two measurements that they both comprise positive and increasing values. Thus it is no surprise that similar plots are presented as proof that the increase of carbon dioxide results from burning hydrocarbons. But is that the case?

Figure 1:
The datasets used here are:

1. Global Emissions: G. Marland, T. A. Boden, and R. J. Andres, "Global, Regional, and National CO2 Emissions," Trends: A Compendium of Data on Global Change (Oak Ridge, TN: Carbon Dioxide Information Analysis Center, 2007); 2005 and 2006 emissions calculated by Earth Policy Institute from energy consumption in BP, Statistical Review of World Energy (London: 2007). 2. Global Carbon Dioxide: Data from Mauna Loa by Scripps Institute of Oceanography, CDIAC, and NOAA/ESRL at, updated January 2008, with long term historical data compiled by Worldwatch Institute from Scripps, ORNL, and IPCC.

A great many people on seeing plots like this immediately claim high correlation and thus assume the two are linked. Unfortunately for them, things are not as they seem, not least because the graph shows two differing measures that cannot be directly compared from that plot. Global emissions are the total amount of carbon added annually to the atmosphere from hydrocarbon usage whilst the global CO2 level is a concentration. It is therefore necessary to calculate the change in CO2 with time so that they can be compared. Changes in CO 2 should be proportional to the quantity of emissions added if there is linkage. An analogy could be the change in colour of the glass of water as successive drops of ink are added and mixed in, provided that the system is not near a saturation point. In the next plot, Figure 2, the annual changes in CO2 (dCO2/dt ) are compared with annual emissions as two time series.

Figure 2:
Now the measures being compared are compatible. Whilst they both have increasing trends, it is now apparent that the changes in carbon dioxide are much more 'noisy'. What might be causing the noise? This is something that will be considered later. Another point worth mention is that in 1964 the annual change in CO2 was negative despite increasing emissions over the period.

Next the correlation between the two is considered and is shown in Figure 3.

Figure 3:
Two linear trends have been fitted; one forced through the origin and the other free. The results are: Y = 0.06083 + 0.0002508*X, R2 = 0.3971 Y = 0.0002614*X, R2 = 0.3964 (A) (B)

Case (A): There is a small positive offset in dCO2/dt at zero emissions. Does this mean that some of the annual increase is natural? That might be one interpretation, however it is well within the anticipated uncertainty and therefore drawing such a conclusion would be ill-advised.

The slope of the trend line indicates the conversion between emissions in Mt C to atmospheric CO2 in ppmv is 3987 Mt/ppmv. This value is not far short of twice that calculated by Trenberth. “Using 5.137 x 10^18 kg as the mass of the atmosphere (Trenberth, 1981 JGR 86:5238-46), 1 ppmv of CO2= 2.13 Gt [2130 Mt] of carbon.” [7]

If it is assumed that about 50% of the anthropogenic CO2 emissions are absorbed annually, then it means that the sink, whatever it may be, is not constant. This is an intriguing proposition since it suggests a sink of increasing capacity, one that has doubled over the last 50 years. Case (B): This yields a slightly different trend with a similar correlation, but a slightly lower 3825 Mt/ppmv value. The same interpretations apply.

Accounting for Temperature

It is important to try to determine the components of the change in CO2 since there are many known natural processes that could affect it, not just humanity burning hydrocarbons. Now the apparent noise on dCO2/dt will be considered since it is probably a clue to the origin of one of the components.

What could be the cause of this variation? It is well known that the solubility of gas in water conforms to Henry's law. This shows that the solubility of CO2 decreases with increasing temperature. Since the Earth surface is about two-thirds covered in water, it would be likely that the concentration of atmospheric CO2 was temperature dependent. This, of course, assumes that in general dissolved minerals do not affect the law unduly. Could the noise on the dCO2/dt signal be partly due to temperature variation?

The annual change in CO2 was plotted on the same time scale as global temperature anomaly from Hadley Climate Reasearch Unit (CRU), CRUTEM3 global means [3], in Figure 4.

By eye, there appears to be some correlation, but to investigate this further dCO2/dt is plotted against temperature anomaly and displayed in Figure 5. If Henry's law was involved then an almost linear relationship should be seen between the two. Henry's law is not linear with temperature over a wide range, but for small changes it approximates quite well and so for the purposes of this analysis the linear assumption is reasonable.

Figure 4:

Figure 5:

A linear trend has been fitted giving the following formula:

Y = 1.141 + 1.494*X, R2 = 0.5290


Whilst the correlation is not spectacular, it is significantly stronger than the CO2 to emissions relationship calculated earlier and thus should not be ignored.

The calculated relationship provides a simple model of the temperature dependence of annual change in CO2 and will be used to try to remove the noise from it. Thus the CRU temperature anomaly will be fed into the dependent term of the formula (C) and subtracted from the measured dCO2/dt for each data point in the time series to produce a version corrected for the effect of temperature. In other words:

Corrected dCO2/dt = Measured dCO2/dt – (1.494*temperature)

The results are shown in Figure 6.

Figure 6:

The correction has reduced the noise amplitude somewhat, but has also made a marked difference to the trend. This suggests that some of the measured increase in atmospheric CO2 is the result of temperature increase and not emissions.

Now the relationship between emissions and dCO2/dt is re-assessed using the temperature corrected data, as shown in Figures 7 and 8.

Clearly the measures look quite different now, but is there a relationship still? Looking at the scatterplot provides the answer.

It should be pretty obvious from this that there is no significant relationship between the temperature corrected dCO2/dt record and emissions, R2 = 0.00537. If it is assumed that the low R2 value is the result of being swamped by noise, then the trend could be considered, but that is virtually flat implying little, if any linkage. This is in stark contrast to the uncorrected relationship of Figure 3.

So again, no linkage is evident between human carbon dioxide emissions and temperature. However, the temperature and carbon dioxide concentration change is explored in more detail later using finer temporal resolutions. Broadly similar results are obtained.

Figure 8:

Figure 7:

Correlation of Detail

In the first section, Basic Correlation, it was seen that both emissions and annual change in CO2 have increasing trends with time. This presents a problem since any two increasing time series will produce a certain level correlation even if there is no causal relationship between them. Therefore in order to try and determine whether there is linkage between these two datasets it is necessary to remove the trend in some way and then reassess them for correlation. There are a variety of methods for doing this but generally the simplest technique works well and that is to calculate annual first order differences in each.

To this end, annual differences of both time series were produced and are plotted in Figure 9. It clearly shows little relationship between the two.

Figure 9:

Producing a scatter plot from this and looking for a linear relationship yields the result given in Figure 11 which demonstrates no significant correlation, since R 2 = 0.021956.

This would tend to suggest that there is another signal present. Could there still be a strong temperature relationship? To find out, the same process was applied with CRU temperature and the results are presented in Figure 12.

Figure 10:
Fitting a linear trend produces the following results:

Y = -0.01075 + 2.523*X R2 = 0.4740

There is a relatively high degree of correlation which tends to indicate that temperature has a far greater influence on changes in atmospheric CO2 levels than emissions.

Figure 11:

Figure 12:

Next a different approach will be taken to try to find a relationship between emissions and change in CO2 by considering the amount of emissions necessary to produce a 1ppmv change in atmospheric CO2. Generally, when a given amount of substance is added to an essentially closed system it would be expected that a proportional increase in concentration should be observed. Thus the ratio of emissions to change in CO2 was calculated for the time series, Figure 13.

It is clear from this that the quantity of emissions to produce a 1 ppmv change in CO2 is highly variable and that the result for 1964 might be regarded an outlier. Since this data point is 'unusual' and can influence trend-fitting, Figure 14 is presented with that point excluded.

In both cases linear trends have been calculated (cyan). Also local polynomial regression (green) smoothed versions have been included which used the following parameters:

Kernel function = Gaussian Degree of polynomial = 2 Bandwidth (Fraction of x range) = 0.1

Returning to the linear trends, the results were:

All data: Y = -78580 + 41.59*X Excluding 1964: Y = -24160 + 14.29*X

There are some major problems here with the assumption that emissions are causing the changes in CO2 which must be considered. The amount of variation is considerable but despite this, there appears to be a gradually increasing quantity of emissions needed for each ppmv of CO2. Assuming that the changes in CO2 observed are due to emissions this suggests that the balance of sinks and sources must be highly variable on short time scales and that the sinks are becoming greater.

Is this the case? Shouldn't some account be made for the natural changes already calculated resulting from temperature variations? To see the effect of taking temperature into account, the previously derived corrected dCO2/dt data was used in the ratio emissions: dCO2/dt calculation, Figure 15.

Figure 13:

Figure 14:

Figure 15:

Figure 16:

Again, 1964 can probably be considered an outlier and so a second graph is presented excluding this point, Figure 16. In both cases the smoothed curves use the same filter settings as stated above and linear trends have been calculated.

All data: Y = -165700 + 86.47*X Excluding 1964: Y = -258500 + 133.05*X

In both cases, if it is assumed that emissions are causing the change in CO2 then the effect is reducing with time, i.e. the quantity of emissions per ppmv CO2 is increasing. Again, this implies a sink increasing with time. Alternatively it could suggest, as has been implied by the earlier analysis, that emissions have very little, if any influence in the atmosphere.

Temperature Lead or Lag?
It is important to determine whether temperature drives CO 2 levels or is the result of CO2 forcing. For this part of the exercise it is necessary to swap to datasets with greater temporal resolution. It is also essential to use the best datasets available and since the satellites provide the best spacial coverage, one of them, in this case UAH temperature anomaly has been selected. There was no particular reason for selecting UAH and similar results should be obtained with others. For the CO 2 record, Mauna Loa Observatory (MLO) data has been selected due to its heritage and prominence. Since CO 2 is described as well-mixed, and MLO is located in the tropics, this choice ought to be reasonable.

The temperature dataset was taken as the reference and the annual change in CO 2 was correlated against it in monthly increments. Both R and R2 for the overlapping data were calculated at each step. R was then used to screen negative correlations from the R2 series. See figures 17 to 19. Figure 19 zooms in on the peak around zero lag to determine the peak and corresponding lag.

Figure 17:

Figure 18:

Figure 19:

From these plots it is very clear that changes in CO2 occur after temperature which is what would be expected for a thermal lag controlling solubility. It is apparent that the correlation is strongest with ocean temperatures, which again suggests a mechanism such as that described by Henry's law.

The main correlation peak is at 5 months and this value will be used in subsequent calculations. Those later peaks are likely to be the result of cyclic phenomena, possibly Thermaline circulation. Some further observations and analysis can be found in the next section, Periodicity in Correlation between Temperature and CO2.

One point that should be mentioned at the juncture is that standard R 2 was calculated and not sample size adjusted R2. Thus there could be some effect due to reduced samples size as the lag time moves away from zero on both sides.

It is important to determine a reference threshold for the significance of the correlation peaks since random variations and other artefacts can produce peaks. To investigate this it was decided that autocorrelations on the two datasets would give an indication of a suitable threshold level. The dCO2/dt record used the same number of points as the temperature record compared to the full full version of the same dataset. This produces slight differences in the pattern either side of zero shift. In the case of temperature, the same dataset is matched and thus there is a mirroring of the pattern about zero shift. See figures 20 and 21.

From these plots a significance threshold of about 0.25 appears to be reasonable. Thus any peak in Figure 17 less than 0.25 needs to be treated cautiously since it could be spurious.

Applying this criterion leaves two peaks, both showing dCO2/dt lagging temperature. The main peak is at 5 months and there is a second at 91 months.

Figure 21:

Figure 20:

Periodicity in Correlation between Temperature and CO2
Earlier in this document the correlation of temperature and dCO2/dt was examined. Two major peaks were identified, with a third smaller one that was considered just below the significance threshold level. This section extends the calculation and appears to identify another peak despite the rapidly reducing data size (Figure 22: ). Clearly, the strength of the correspondence will be compromised and it is more noisy, but it does not appear to be entirely spurious. The fairly large peak after 300 months lag is almost certainly an artefact since it is much narrower and more ragged than any of the others. Not only that, but the dataset is very short by that stage of the calculation, less than 20% of the maximum.

Figure 22:

Taking the lag values of the identified peaks (months) and difference between them:

Lag (months)


91 86

184 93

271 87

Difference (months) N/A

The differences are fairly similar suggesting a possible oscillatory mechanism. Assuming this is the case then the mean period is 88.7 months with SD of 3.8 months (7.39 ± 0.32 years). Is this periodicity the result of an ocean cycle? There is certainly literature suggesting exactly that; the North Atlantic Oscillation (NAO) has a similar periodicity to that discovered here. The following literature examples report comparable findings:

1. The 7.7-year North Atlantic Oscillation, E. D. Da Costa and A. Colin de Verdiere Q. J. R. Meteorol. Soc. (2002), 128, pp. 797–817. 2. Statistics of North Atlantic Oscillation Decadal Variability, Jason Goodman, Program in Atmospheres, Oceans and Climate Department of Earth, Atmospheric and Planetary Sciences, Massachusetts Institute of Technology. 3. Rainfall Variability in Southern Spain on Decadal to Centennial Time Scales, F.S. Rodrigo et al. International Journal of Climatology Int. J. Climatol. 20: 721–732 (2000). 4. Fluctuations in North Atlantic Hurricane Frequency, J. B. Elsner, A. B. Kara, and M. A. Owens, Department of Meteorology, The Florida State University, Tallahassee, Florida, American Meteorological Society 1999. 5. North Atlantic Oscillation Dynamics Recorded in Greenland Ice Cores, C. Appenzeller et al, 16 October 1998 Vol. 282 Science. 6. An Analysis of the Variability of the North Atlantic Oscillation in the Time and the Frequency Domains, D. Pozo-Vazquez et al. International Journal of Climatology Int. J. Climatol. 20: 1675–1692 (2000).

From this analysis, it would seem reasonable to draw the conclusion that rather than the oceans being a sink of CO2 as is generally accepted, at present they are actually, on balance, a source driven by ocean temperatures which in turn are related to NAO.

Purely as an observation, it is a remarkable coincidence that the periodicity found by this correlation of 7.39 ± 0.32 years is very close to the mean atmospheric residence time of 7.6 ± 2.5 years calculated from all the reported values from Appendix 4. Whether it is any more than coincidental is beyond the scope of this study.

Accounting for Temperature with Greater Accuracy
From the previous section the lag between UAH temperature and dCO2/dt has been found and will be used to realign the two datasets such that a relationship between the two can be determined more accurately. Whist there is a slightly better correspondence between ocean temperature than the other two, this analysis will use global temperature. In practice, because two thirds of the Earth's surface is ocean, it makes no difference to the conclusions. Plotting the datasets shows the similarity between the two as evidenced by the lagged cross correlations, Figure 23. Shifting dCO2/dt back in time to align with temperature by the amount calculated from the main correlation peak yields the time series graph in Figure 24. Then the time shifted dCO2/dt was plotted against temperature and the linear relationship between the two computed, Figure 25. This yields the following fit equation: Y = 1.548 + 1.940*X, R2 = 0.395 Thus a model of dCO2/dt as a function of temperature is produced and feeding back into it the UAH global data creates a dCO2/dt time series which is the result of temperature variations. Subtracting this signal from the measured dCO2/dt gives the error signal as shown in Figure 26. This error (residual) ought to reproduce the emissions data, assuming the overall model is correct. A linear fit has been added and it can be seen to have almost no slope; actually it is slightly negative. It could be argued that only the temperature dependant term should be used in the dCO 2/dt created from temperature rather than the full expression. If that premise is accepted, it adds another problem; that being that the dCO2/dt trend retains the same flat gradient and in order to obtain a forced match to emissions another source of CO2 with a complimentary trend would need to be identified. For completeness, Figure 27 shows the temperature corrected dCO2/dt without removing the constant term along with emissions time series. The annual dCO2/dt and annual emissions are shown in Figure 28. Cross-plotting the data presented in Figure 28 shows no correlation between the two parameters as illustrated by Figure 29.

Figure 23:

Figure 24:

Figure 26:

Figure 25:

Figure 28:

Figure 27:

Figure 29:

Another important aspect of this analysis is that compared to the version based on CRU temperature data, there is a significant difference in the linear models; the slopes are 1.940 and 1.494 respectively. Whilst there are several possibilities for the discrepancy, probably the most likely is that there is a warm bias in the CRU data. This is supported by the fact that it has a lower slope value and that it produces a downward trend in the corrected version. It seems even more likely in light of the recent controversy regarding informally released documents from CRU at the University of East Anglia that had been compiled under the Freedom of Information Act.

This suggests that all the increase in atmospheric carbon dioxide could be accounted for by changes in ocean temperature.

An additional set of graphs are contained in Appendix 3: Additional UAH Temperatures & dCO2/dt relationships. They show the comparable UAH land and ocean temperatures and MLO dCO2/dt data and relationships.

Cumulation with Half-life & Absorption Parameters
Many of the analyses by other researchers, as noted earlier, use a cumulation method applied directly to fossil fuel emissions. These processes generally add scaling and half-life parameters, with varying definitions and terminology.

The time dependent parameter (half-life, residence time or lifetime) has been defined in several ways and consequently has multiple names. [6] Irrespective of its name, it is basically a measure of how long it takes from a given input of CO2 to disappear from the atmosphere. Thus the longer it is, the more build up of CO2. Generally e-folding time and half-life are used. E-folding is about 44% longer than half-life. [14] [15] In this study the terms half-life and residence time will be considered interchangeable, except where noted otherwise.

It is interesting to note that the IPCC states “Carbon dioxide does not have a specific lifetime because it is continuously cycled between the atmosphere, oceans and land biosphere and its net removal from the atmosphere involves a range of processes with different time scales”. Yet in the notes at the bottom of page 34, AR4 WG1 Technical Summary this is written: “The CO2 response function used in this report is based on the revised version of the Bern Carbon cycle model used in Chapter 10 of this report (Bern2.5CC; Joos et al. 2001) using a background CO2 concentration value of 378 ppm. The decay of a pulse of CO2 with time t is given by a0 + Σ ai . e ^(-t/τi) [summation over i = 1 to 3] where a0 = 0.217, a1 = 0.259, a2 = 0.338, a3 = 0.186, τ1 = 172.9 years, τ2 = 18.51 years, and τ3 = 1.186 years, for t < 1,000 years.”[16] So a model has been derived and has a dominant lifetime of 172.9 years. This is an e-folding lifetime and so its corresponding half-life is 120 years. The relevance of this will be seen later.

The pulse model can therefore be written:

CO2 (t) = CO2 (t=0) * [0.217 + 0.259*e^(-t/172.9) + 0.388*e^(-t/18.51) + 0.186*e^(-t/1.186)]

Where t = time from the CO2 input in years.

Because the two shorter lifetime terms have little effect on the build up of CO2 in the atmosphere they can be ignored and the equation becomes:

CO2 (t) = CO2 (t=0) * [0.217 + 0.259*e^(-t/172.9)]

Within the context of the datasets, the e-folding time of 172.9 years is essentially infinite. Thus the expression simplifies to:

CO2 (t) = CO2 (t=0) * [0.217 + 0.259]

CO2 (t) = CO2 (t=0) * 0.476

This is probably where the often cited (provenance unknown) “about half of the emissions stay in the air permanently” originates.

Since the UN IPCC uses a residence time greater than 100 years, the small effect of differences between definitions of e-folding and half-life is generally insignificant.

Any cumulation-based process applied to a series of positive numbers will tend yield an upward trending line, as shown by example in .

The process is can be extended to show the effect of scaling and application of half-life on cumulation.

Using the simplest form of such a function it is demonstrated that this approach can yield misleading results.

The basic form of the discretised equation for emissions in the atmosphere is:


A = Absorption (scaling) factor T = Half-life parameter (Residence term) Totalt = Total emissions at time = t New = Newly added emissions Totalt-1 = Previous total emissions

Here is an example of how a dataset can be spuriously correlated with CO2 concentration through the inappropriate use of the above process (cumulation). This section was written as an article for instructive purposes and thus is in less formal language. NB: The equation is basically the same although the naming is slightly different.


Is the Met Office to Blame for the Rising CO2 Levels?

Recently I was trying to explain for the millionth time that the cumulation maths process can make almost any positive data set with an upward trend look like the increase in CO2 within the atmosphere. As usual I presented a few examples using simple synthetic datasets, but somehow they don’t seem to be sufficiently compelling. What might be better than a simple example? How about an amusing one?

I was browsing the news and came across a bulletin claiming that the new Met Office computer was extremely energy hungry (not to mention expensive), but that it would improve the predictive skill. So I had a quick look at the Met Office website and fell across a plot comparing forecast accuracy year on year. It was hardly surprising to find it is a positive dataset with an upward trend. Just the kind of data required for a little demonstration. It was found here: Sourced 9/8/2009

Figure 30:

I borrowed the graph and extracted the values:

Figure 31:
For comparison I have plotted the annual change of CO2 in the atmosphere on the same chart:

Figure 32:

What can you say? Hardly a good match, but then why would it be? Anyway, now we have two datasets to which the same cumulation method will be applied. Units of Met Office data (days) will be replaced by CO2 ppm. Each will have a different pair of variables. The variables are a scaling factor and a half-life. This simple function was used: CO2 = (Old*EXP(-LN(2)/Half_life))+(New*Scale) Where: Old = the amount of CO2 left after the last year New = quantity of CO2 added this year From the model, time series were built and are shown in the next graphic. NB: For expediency, the two parameters for each were adjusted to give a reasonable visual fit. They were not optimised.

Figure 33:
It can be seen that both datasets produce a reasonable fit to the MLO CO2 measurements. Just for the record, the following values were used:

Met Office forecast: Scaling = 1.4, Half-life = 9.4 years Emissions: Scaling = 0.57, Half-life = 125 years

Coincidentally, the half-life of 9.4 years for the Met Office set is within the range of reported values for the residency of CO2 in the atmosphere (See Appendix 4), but it is exactly that, a coincidence. As it is difficult to see the variation between the time series in the above, time series of the errors between the two sets and MLO history were created:

Figure 34:
Well, what can I say? That is a pretty good comparison. Could this be damning evidence that the Met Office has caused the rising levels of CO2 in the atmosphere? Of course not! What it does show is that if cumulation is used to create time series, it must be supported by strong evidence for, and tight constraints on any variables, ideally with a valid physical mechanism.

Is the Met Office to Blame for the Rising CO2 Levels? Unfortunately no, but there is little doubt it has contributed.


Thus the importance of using correct parameters in this mathematical procedure is paramount. Unless there is substantial justification for the parameters and approach, the validity of the model must be questioned. This is most certainly the case here. An arbitrary absorption factor and a grossly unrealistic half-life have been used. This can be seen in the graph presented in Appendix 4 based upon a compilation by Tom V. Segalstad. His data gives a mean CO2 residence time of 7.5 ± 2.5 years.

To summarise, the cumulation of emissions with the parameters set out by the UN IPCC is nothing more than a curve fitting exercise and does not represent a physical model of system. Almost any allpositive data with a slight upward trend can be forced to closely match the Mauna Loa atmospheric carbon dioxide record by application of this mathematical process.

There is no evidence to support the claim of CO2 residence times of 100's of years or longer. It is based upon the assumption that human CO2 emissions accumulate in the air and that there is no naturally changing background of CO2. These false assumptions are being used to support the global warming hypothesis and the so-called 'historic responsibility' of industrialised nations in the imagined scare story. More on this topic can be found in Appendix 5.

Isotopic Changes 13C/12C
It is often stated that isotopic 13C/12C changes show the signature of ‘fossil’ fuel combustion, but is there an alternative explanation? There seems little dispute that CO2 is cycled in and out of the ocean surface over short time scales. The periods for average bulk transitioning could be of the order of hours, are highly variable, and are dependent upon local conditions. This is in contrast to the claimed residency time that has been dealt with elsewhere in this study which is essentially a global parameter. It is well documented in the literature that fractionation occurs and carbon dioxide passes through the air/seawater interface. It is known that the ocean surface is a dynamic system whereby CO2 passes to and fro from the ocean to air over short periods of time, and with the exchange of gas in each direction, there is an isotopic modification of the carbon dioxide composition. The dominant fractionation process is from ocean to air and so any global average out-gassing of the ocean would create a change in atmospheric 13C/12C ratio. This process comprises several established physical and chemical mechanisms, could explain the changes in isotopic carbon ratios equally well, and probably better than, the assumption that they are due to the burning of ‘fossil’ fuels by humans.

From the analyses shown, it can be seen that there is no evidence to suggest that carbon dioxide from the burning of hydrocarbons has any influence on atmospheric levels. There is significant correlation of atmospheric CO2 with ocean temperature which might be explained by Henry's law of gas solubility in water. It is also shown that changes in CO2 occur after changes in temperature, and in particular ocean temperature. An apparent periodicity in the correlation may suggest a link to ocean circulation patterns. The idea of natural saturating CO2 sinks as suggested by the UN IPCC has been shown to be flawed because it simply is not supported by the data. By example it has been shown that cumulation of all positive time series with a slight upward trend can be made to closely match the Mauna Loa atmospheric CO2 record even when there is no linkage between datasets. This process cannot be used to demonstrate a relationship between anthropogenic releases of CO2 from burning hydrocarbons as it is inappropriate and leads to the spurious long atmospheric lifetimes for CO2 used by the IPCC. Although not dealt with in detail in this work, isotopic changes in atmospheric carbon dioxide (13C/12C ratio) could be explained by dynamic fractionation taking place at the air/ocean interface. Global average out-gassing of carbon dioxide would inherently crate a change in atmospheric isotopic ratio. Since there is no significant linkage between atmospheric CO2 and burning hydrocarbons; and an incorrect longevity has been assigned to CO2 emissions; and an invalid relationship between atmospheric CO2 and global temperature used in climate models the results from them, i.e. future climate scenarios, are without doubt worthless. Thus all policies based upon the proclamations of the UN IPCC are simply ideology, have no scientific grounding, and should be either ignored or reevaluated.

Data & References
1. Global Emissions: G. Marland, T. A. Boden, and R. J. Andres, "Global, Regional, and National CO2 Emissions," Trends: A Compendium of Data on Global Change (Oak Ridge, TN: Carbon Dioxide Information Analysis Center, 2007); 2005 and 2006 emissions calculated by Earth Policy Institute from energy consumption in BP, Statistical Review of World Energy (London: 2007). 2. Global Carbon Dioxide: Data from Mauna Loa by Scripps Institute of Oceanography, CDIAC, and NOAA/ESRL at, updated January 2008, with long term historical data compiled by Worldwatch Institute from Scripps, ORNL, and IPCC. 3. Hadley CRU Global Temperatures: CRUTEM3 global means: 4. MLO CO2 data: File Creation: Mon July 13 09:08:40 2009. 5. UAH Temperature Anomaly Data: Dated 4-8-2009:

6. Tom V. Segalstad, Mineralogical-Geological Museum, University of Oslo, Sars' Gate 1, N0562 Oslo, Norway. Carbon cycle modelling and the residence time of natural and anthropogenic atmospheric CO2: On the construction of the "Greenhouse Effect Global Warming" dogma. 7. Carbon Dioxide Information Analysis Center - Frequently Asked Global Change Questions (Trenberth, 1981 JGR 86:5238-46) 8. The 7.7-year North Atlantic Oscillation, E. D. Da Costa and A. Colin de Verdiere Q. J.

R. Meteorol. Soc. (2002), 128, pp. 797–817 9. Statistics of North Atlantic Oscillation Decadal Variability, Jason Goodman, Program in Atmospheres, Oceans and Climate Department of Earth, Atmospheric and Planetary Sciences, Massachusetts Institute of Technology 10. Rainfall Variability in Southern Spain on Decadal to Centennial Time Scales, F.S. Rodrigo et al. International Journal of Climatology Int. J. Climatol. 20: 721–732 (2000) 11. Fluctuations in North Atlantic Hurricane Frequency, J. B. Elsner, A. B. Kara, and M. A. Owens, Department of Meteorology, The Florida State University, Tallahassee, Florida, American Meteorological Society 1999. 12. North Atlantic Oscillation Dynamics Recorded in Greenland Ice Cores, C. Appenzeller et al, 16 October 1998 Vol. 282 Science AO%20&%20Green%20Ice%20Cores.pdf 13. An Analysis of the Variability of the North Atlantic Oscillation in the Time and the Frequency Domains, D. Pozo-Vazquez et al. International Journal of Climatology Int. J. Climatol. 20: 1675–1692 (2000) 14. 15. 16. 17. U. Siegenthaler & K.O. Monnich: 13C/12C Fractionation during CO2 Transfer from Air to Sea:

Appendix 1: Examples of Simple Cumulation on Time Series

Here are some further examples of simulated datasets put through the basic cumulation process. In all these cases the absorption factor =1 and the half-life is infinite.

Figure 35:

Figure 36:

Figure 37:

Figure 38:

Figure 39:

Appendix 2: Changes to MLO CO2 Record
Derek Alker, recently spent much time and effort looking over the MLO carbon dioxide measurement technique and data. He kindly supplied a number of datasets of differing dates. The following graph was constructed by simply taking the differences between sets, referencing to the 17 December 2008 release.

Figure 40: Differences in MLO data relative to 17 December 2008

It is believed that the major adjustments between 2000 and 2008 are documented. However, one thing that draws question is how the apparently random differences mostly prior to 1973 were eliminated in later datasets?

Appendix 3: Additional UAH Temperatures & dCO2/dt relationships

Figure 41:

Figure 42:

Figure 43:

Figure 44:

Figure 45:

Figure 46:

Appendix 4: Compilation of CO2 Atmospheric Residence Time

Figure 47: Carbon dioxide residence time by author & publication. Data compiled by Tom V. Segalstad. Figure 48:
Data: Tom V. Segalstad, Mineralogical-Geological Museum, University of Oslo, Sars' Gate 1, N-0562 Oslo, Norway.

Mean CO2 residence time = 7.6 ± 2.5 years.

Appendix 5: Bern Carbon Cycle Model Bern Carbon Cycle Model

Many of the climate models projecting into future scenarios use the Bern Carbon Cycle model as highlighted by the IPCC. Clearly, the IPCC considers this to be an important piece of their evidence and is apparently the only one where more than a passing citation is included. It is unclear as to why this particular model should hold so much prominence and thus some research has been done to try and understand the reason for this. Indeed, the Bern Carbon Cycle model (2.5) [1] is cited in AR4 WG1 referring to its use in chapter 10 [3] of the report [2]. An extract of this text, which is found on page 34 of WG1 can be found in Error! Reference source not found.51. This particular piece of text provides the decay response of CO2 with time. Its importance is in the fact that it describes the effective residence time of CO 2 in the atmosphere and should be in accordance with estimates from empirical analyses. It assumes that nature distinguishes between CO2 resulting from hydrocarbon burning and that from biological metabolism or geological expulsion. Thus it applies only to CO2 from ‘unnatural’ anthropogenic activities, in other words burning so-called fossil fuels. This essentially assumes that the natural world can discriminate between CO2 from fossil fuel combustion (FFC) and natural sources. This, of course, is not the case and consequently is a shortcoming of the model because it needs to be able to accommodate fluctuations in natural CO2 exchange. However, that is not accomplished by the model since any input will result in an increased atmospheric level of CO2 as will be explained later. Having noted this fact, the remainder of this document will assume that the model only applies to CO2 from FFC. In essence the formula presented is a multiple (triple) exponential decay expression with an additional constant factor. It is supposed to represent the decay of an input of CO2 to the atmosphere over time periods less than 1000 years. The expression can be written as: CO2(t) = a0 + a1*exp(-t/T1) + a2*exp(-t/T2) + a3*exp(-t/T3) Where: a0 = 0.217 a1 = 0.259, T1 = 172.9 years a2 = 0.338, T2 = 18.51 years a3 = 0.186, T3 = 1.186 years t = time after injection of CO2 in years Equation 1.

Whilst there is considerable controversy about the decay constants, in particular T1 and T2, there is a far more fundamental issue. The term a0 has no associated decay constant which means that there is an intrinsic decay constant equal to infinity. Thus for any given input of CO 2, 21.7% of it will remain permanently in the atmosphere. This means that any input of CO2 from fossil fuel combustion will result in an accumulation, i.e. Increase in concentration. Indeed, if it is to be believed, any random surge or spike in natural input into the atmosphere from any source, natural or not, will result in a permanent increase in atmospheric CO2 concentration. There is no provision for a drop in the long term (centuries) CO2 concentration within the model. Equations are all well and good for those familiar with maths, but for the average person a graph is far more useful in understanding the function. To that end, the function has been plotted, as shown in Error! Reference source not found.49. From this the asymptotic nature of the curve can be seen which indicates the permanent accumulation of CO2 into the atmosphere granted by formula.

Figure 49: Bern 2.5 CC CO2 pulse decay formula. It shows the proportion of remaining after a unit input of CO2.

This is not shown in Joos et al 2001. However, it is interesting to note that a similar plot is provided in supporting information for Solomon et al. 10.1073/pnas.0812721106: The Solomon et al graphic is presented in the appendix, Error! Reference source not found.50. It is notable that its asymptotic level is much greater at about 0.45 rather than the 0.217 from IPCC WG1. The second and third terms of the Bern CC model (a2, T2, a3, T3) account for the rapid loss of about half the anthropogenic CO2 emissions, leaving the other half as accumulation in the atmosphere. Hence, there is an assigned “absorption factor” of around 0.5 which is regularly cited. A question should be asked at this juncture and it is; why should there be an absorption factor? If CO2 sinks have a limited capacity and become saturated or near saturated then that capacity will be a constant or near constant, not a proportion of FFC. Returning to Equation 1, the ascribed dominating residence time and absorption factor between FFC emissions and atmospheric CO2 can be calculated from a curve fitting exercise without recourse to any physical process. This method is readily and particularly applicable to the instrumental CO2 data and emissions records yielding values similar to the dominant components of the Bern CC impulse model. However, any all-positive dataset with a slight upward trend can be forced to fit the longest instrumental CO2 record which comes from Mauna Loa (MLO) with a high degree of accuracy by choosing two variables appropriately. These variables do not necessarily have any physical significance whatsoever as was shown in the example given here: Is the Met Office to Blame for the Rising CO2 Levels? Note that the e-folding lifetimes used in Bern CC are about 44% longer than the equivalent half-life. Thus the half-life values used in the example should be adjusted to be comparable, i.e. an e-folding time of 172.9 years is a half-life of 120 years.

But where did this all start? These long-decay impulse responses appear to largely originate from the work of Joos and cohorts. Back in 1996 the impulse function [Figure 52] was apparently quite consistent with estimates in the literature, as compiled by Tom V. Segalstad [Appendix 4: Compilation of CO2 Atmospheric Residence Time] which represent responses from incremental loading of the atmosphere with carbon dioxide. By 1999 Joos et al changed tack and presented impulse functions based upon an instantaneous doubling of the atmospheric concentration of CO 2 as shown in Figure 53. An instantaneous doubling of CO2 has not happened and is extremely unlikely to ever happen, unless there is some catastrophic event such as a massive meteor impact. Indeed, CO2 concentrations have not doubled since the purported ‘pre-industrial’ levels and it may not happen within the lifetimes of most people living today. As such the residence times used in the global emissions models are invalid, as are the model outputs.

The CO2 impulse decay models presented in AR4, and others published by closely associated entities, are inconsistent and apparently lacking a solid scientific basis. These models, and the processes that lead to them, and their use, should be fundamentally questioned. Physical and empirical evidence must be presented to support all their underlying mechanisms and assumptions. It is suggested that the unrealistically large residence times and substantial constant components of these models have been introduced, without scientific foundation, for political purposes. These parameters are the basis for the creation of an artificial legacy applicable to the richest (most industrialised) nations. Using this alleged historic responsibility as a political lever has lead to inappropriate carbon taxation systems and mechanisms that will adversely affect those states pursuing them.

1. Joos et al, 2001 2. Solomon, S., et al, 2007: Technical Summary. In: Climate Change 2007: The Physical Science Basis. Contribution of Working Group I to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change. 3. Meehl, G. A., et al, 2007: Technical Summary, Chapter 10, Global Climate Projections. In: Climate Change 2007: The Physical Science Basis. Contribution of Working Group I to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change.


Figure 50: Graph from supporting information, Solomon et al. 10.1073/pnas.0812721106

Figure 51: Extract from IPCC AR4 WG1.

Figure 52: Extract from Joos et al 1996: Pulse response functions are cost-efficient tools to model the link between carbon emissions, atmospheric CO2 and global warming.

Figure 53: Extract from Joos et al 1999: Correcting the carbon cycle representation: How important is it for the economics of climate change? Impulse response to Instantaneous Doubling of CO2.