Treatment of Uncertainties for National Estimates of Greenhouse Gas Emissions

5. Discussion
   

As can be seen from Figures 1a and 3a, the predicted uncertainties in emissions in 1990 and 2010 are both approximately 4%, a value much greater (roughly three times) than the predicted increase in emissions over this time ¹². The distribution of changes in Figure 3 illustrates very clearly that no significant change in emissions can be anticipated in 2010 compared to 1990 on the current bases of the activity information given by DTI [13] and the uncertainties in the parameters used to evaluate the emissions inventory itself. Figures 1b and 3b show the results for CO2 emissions from fuel combustion alone. The uncertainties are significantly smaller, 2.1% in 1990 and 1.2% in 2010. This reflects the fact that emissions from fuel combustion are related to the carbon content of the fuel that is well known and easily measured.

In the context of the sinks (Figures 2 and 4), firstly they are of much lower magnitude than the emissions (~3%), insignificant in comparison. Being based on land use changes, they are subject to much larger uncertainties, amounting to ~35-45% of the expectation. In addition, it should be noted that there is expected to be a increase in removal of carbon dioxide, with expected values in 2010 amounting to about 150% of those in 1990. The greater uncertainties in the sinks in 2010 compared to those in 1990 means that the distribution of possible increases in removals has a long tail, reflecting the fact that the increased removal of carbon dioxide between 1990 and 2010 could, with a significant probability, actually be much greater (up to a factor of two or three) than would be apparent from a ratio of the corresponding mean change of 50%.

5.2 Methane

Predicted uncertainties in emissions in 1990 and 2010 (Figures 5 and 6) are in the region of 15-20%. This is larger than for carbon dioxide, and reflects a lower confidence in the quality of the data used in calculating the inventory, and possibly also, greater inherent variations in parameter values themselves (see Section 3.2). The mean inventory in 1990 of 4300 kt falls to 2700 kt in 2010. This reduction of ~36% is much more significant than the magnitude of the uncertainties in emissions in either year. The distributions of emissions in 1990 and 2010 shown in Figure 6 can be seen to be practically non-overlapping, illustrating the predominance of the anticipated decrease over component uncertainties.

5.3 Nitrous Oxide

Figures 7 and 8 display the uncertainties in emissions of nitrous oxide. Emissions in general can be seen to fall, from just over 200 kt in 1990 to just over 140 kt in 2010. A number of changes contribute to this change, dominated by a virtual cessation of emissions from adipic acid manufacture (initially a very significant source), and falls in most other sectors. Emissions from fuel combustion (which form about 5% of the total in 1990) increase by a factor of four over the time period 1990-2010, making up about 27% in 2010.

Between 1990 and 2010 the uncertainties in emissions increase modestly, from just over 230% to just over 290%. This follows from the changes outlined above. Table 4 shows that the uncertainties in the emission factors from combustion (dominated by power generation and road traffic) greatly exceed those from adipic acid manufacture.

Following on from the above, Figure 8 shows that the distribution of possible emissions in 2010 is noticeably broader than in 1990. The long tails of both distributions are both due to the great uncertainty in emission from agricultural soils. Moreover, the relative contribution of the agricultural source (which is the most uncertain) increases due to the fall off in emissions from adipic acid manufacture.

The distribution of changes in emissions from 1990 to 2010 shows clearly that in nearly all circumstances, the emission of nitrous oxide is expected to fall between 1990 and 2010. The fine structure visible on the distribution itself is due to the fact that a subtraction is being conducted between distributions for 1990 and 2010 with very broad tails. Owing to the relatively low probability densities over large parts of these distributions, and because of the quantisation of the log normal distributions themselves ¹³, statistical 'noise' is more apparent for this calculation than for the others presented in this report.

5.4 Halocarbons and Sulphur Hexafluoride

Figures 9 and 10 display the uncertainties in emissions of hydrofluorocarbons (HFCs). Emissions can be seen to rise substantially from just over 1000 t in 1990 to about 3200 t in 2010, the result of a number of concurrent increases and decreases in different sub-sectors.

Uncertainties in emissions of HFCs are predicted to be about 25% in 1990, being dominated to above by the uncertainties in the manufacturing losses of HCFC22. This value is itself dominated by uncertainties in the emission factor during manufacture (~25%); the uncertainty in production is estimated to be about 2%. In 2010 the emission from HCFC is predicted to become insignificant due to the introduction of abatement measures and a 42% fall in production. The breakdown of estimated releases in 2010 (as the average of the high and low scenarios) is as follows. The emissions here are dominated by a number of new sources which either did not exist in 1990 or were trivial.

The largest single source of HFC in 2010 is refrigeration, which is the sum of 9 different product types. Although some of these sources are larger than the others there is not a large disparity between most of them. The dominant release mechanism seems to be leakage during the product lifetime. The uncertainties in the individual leakage rates are assumed to be around 20-30%. With aerosols, the emission is set to be equal to consumption. This assumes that aerosols are used within a year of production. Hence the leakage rate is ~1, with negligible uncertainty. With foams, the production losses are of a similar magnitude to the leakage losses. Uncertainties were judged to be higher, at around 50% for PFC and polyurethane foam, but lower (~30%) for polystyrene foam where loss rates were significantly higher.

The total predicted emission of HFCs in 2010, therefore, comprises the sum of a number of sources, many of similar magnitude. As the individual parameters are not linked in any way, this has the effect of producing a total emission with a lower uncertainty than any of the component sources, in this case about 8% ¹⁴. From Figure 10 it can be seen there is no discernible overlap between the probability density functions of emissions in 1990 and 2010. Hence the increase in emissions during this time period is much more significant than the uncertainties in either year.

Figures 11 and 12 display the uncertainties in emissions of perfluorocarbons (PFCs). In contrast to HFCs, the emissions can be seen to fall substantially over this time, from about 300 t in 1990 to just over 100 t in 2010. In 1990 emissions were dominated by aluminium production (97%). The error in the emission from aluminium smelting was estimated at about 20%, a value based on manufacturers' data.

In 2010, emissions from aluminium manufacture are predicted to fall due to the introduction of abatement measures. In 2010 the uncertainty in aluminium production is still assumed to be 20%. The uncertainty in solvent use is low (~10%) as emissions equal consumption (i.e. emission factor equals 1, see earlier). The other sources depend on leakage or sporadic usage so their uncertainties were judged to be higher (30-50%). In spite of this, the overall uncertainty in emissions in 2010 is predicted to be lower, at 13%. As for HFC, this reduction in uncertainty between 1990 and 2010 appears to be driven by one dominant emission source initially evolving into five significant, but unrelated components over the two decades. The change in emissions shown in Figure 12 reflects the fact that there is no overlap between the emissions density functions in 1990 and 2010, and the decrease in emissions during this time period is much more significant than the uncertainties in either year.

Figures 13 and 14 display the uncertainties in emissions for sulphur hexafluoride (SF6). Emissions can be seen to rise substantially from 24 tonnes in 1990 to 43 tonnes in 2010. There are only two significant sources of SF6, cover gas in magnesium production and use in electrical equipment as an insulator.

The cover gas emission is equal to consumption. This was judged to have an uncertainty of about 15%, based on the degree of rounding of the data. Electrical emissions depend on leakage, which was judged to have an uncertainty of about 40%. The estimates suggest that there is no change in the overall relative uncertainties in emissions between 1990 and 2010, and these are around 13%. This seems reasonable as the contribution of cover gas to the total only changes slightly from 87% in 1990 to 83% in 2010. From Figure 14 it can be seen that there is no discernible overlap between the probability density functions of emissions in 1990 and 2010. As for HFC and PFC, therefore, the change in emissions (here an increase of about 80%) is much more significant than the uncertainties in either year.

It can be noted that the (best estimate) total impact of the emissions of gases in this sub-section falls by a factor of about three over the time period 1990 to 2010. The opposing contributions of the individual gases translate into a net fall in the overall emission uncertainty, from 20-25% to under 9%.

5.5 Total Emissions (GWPs)

As can be seen from Figures 15 and 16, the total warming potential of the emissions of the six greenhouse gases considered in this study fall by about 6% between 1990 and 2010. This is much lower than the predicted uncertainties in the overall emissions at these times (just under 20%), although the fall is still statistically significant. As can be seen from Figure 17 and Table 7, the predicted fall in total global warming potential does not arise primarily from any single gas. Figure 16 and 17 show that the distribution of possible change in emissions ranges from around -2% to about -10%.

5.6 Main Sources of Uncertainties

Owing to the very large number of contributions to the overall uncertainty in emissions and sinks of greenhouse gases in the UK (to each gas separately as well as to the weighted totals), the statistical procedure employed was only able to identify the largest contributions in this regard. These are given in Table 7. As noted earlier (Section 4.3), the uncertainty in no single parameter influenced the uncertainties in the emissions of carbon dioxide, methane, nitrous oxide or the aggregate total significantly. Very many sources of uncertainty contribute to the overall effect for these gases, and none of them dominate as their individual effects are 'diluted'. For HFCs, PFCs and SF6 there are fewer sources (and hence model input parameters), particularly in 1990, and so some of them turn out to be significant sources of uncertainty as can be seen. The large figures in 1990 indicate that these are the only significant sources of uncertainty in this year. The different major contributors in 2010 reflect the change in the composition and sources of the tabulated gases over the period 1990 to 2010.

The uncertainties in the values of a number of the identified sensitive parameters in Table 7 may also be seen to contribute significantly to the uncertainties in the predicted changes in emissions between 1990 and 2010. For such changes, variations in parameters can act in either sense, so that in some circumstances (e.g. declining parameter significance over time) a decrease in the value of a parameter between 1990 and 2010 (i.e. an increased fall, or decreased rise) may lead to an increase in the change between these two times (i.e. a decreased fall, or increased rise), giving rise to a negative rank correlation coefficient.