Treatment of Uncertainties for National Estimates of Greenhouse Gas Emissions

A2 Uncertainty Evaluation: Other Specific National Approaches

A number of approaches adopted in other countries to evaluate uncertainties in emissions estimates were reviewed briefly.


It is reported [72] that various levels of uncertainty in emission estimates are associated with different sections of the inventory as a whole, although no information is given on the methodology used to estimate such uncertainties. It is stated that difficulty in obtaining accurate statistics and relevant data greatly hinders the process of estimation, be it in qualitative or quantitative form. Notwithstanding, it is recognised that the emission estimates for transport are likely to be subject to a relatively small uncertainty (~10%), whereas others, such as the emissions due to land use change and forestry, are subject to a much larger uncertainty (about a factor of two). Research is underway to refine estimates of emissions in areas believed to be subject to the greatest uncertainties.


Uncertainties in emission estimates of greenhouse gases are a major concern to the Canadian authorities [73]. Although it is recognised that there are many causes of uncertainties, most are believed to be due to the following:

The document refers to the use of mean and standard deviations of expert estimates of sectoral emissions. A document summarising the full methodology for assessing uncertainties [61] has been requested for use in this work. The overall uncertainties for the three main greenhouse gases were developed using a stochastic (probabilistic) model, and are estimated to be about 4% for CO
2, 30% for CH4 and 40% for N2O. Uncertainties in individual sectors are acknowledged to be much higher than these figures. Overall uncertainties in the emissions of CO2, which dominate, are believed to be relatively low.


The approach adopted by Denmark in assessing uncertainties in greenhouse gas inventories is not given in any meaningful detail [74]. Uncertainties in emission estimates are recognised as arising from two sources: uncertainty in the statistics and uncertainty in the emission factors used. The emission factors are based either on calculations, as is the case for carbon dioxide (from energy and carbon content), or more directly as inferred from measurements. It is recognised that a further uncertainty arises from a lack of completeness in the available data.

The greatest uncertainties in estimates arise for the non-methane volatile organic compounds, methane, and nitrous oxide, with an uncertainty factor of about 2. With the carbon monoxide and oxides of nitrogen inventories, the uncertainty is assumed to be less than 30-40%. The carbon dioxide uncertainty may be as low as 1-2%. No information is given on how these numerical values were derived.


It is widely recognised [75] that emissions data are subject to considerable uncertainty, due to lack of information on emissions-causing processes, and (more importantly) a lack of data on the extent of certain emission-relevant activities. Much effort has been devoted in Germany to composing scenarios of such future activities.

Emission factors for non-CO
2 sources have been based on measurements made under defined conditions, although the number of measurements is regarded as inadequate. This also applies to non-energy-related emissions. Emissions from combustion-related activities are considered to be much more reliable.

Beyond stating that research is being focused on improving the quality of the information used in composing national estimates, the information given in the above reference does not permit the reader to deduce in any way the method or methods used to assess the level of uncertainties in inventory estimates of greenhouse gases in Germany.

The Netherlands

Numerical uncertainty estimates for the inventories of greenhouse gases are given [76], although it is not stated how these were derived. A combination of expert opinion and some numerical manipulation is therefore suspected.

In the recent document outlining the Netherlands' emissions data [77], the percentages given above are reproduced without comment, together with the more qualitative indicators of data quality promoted by the IPCC (see Section 2.1.2(b)).


The official communication [78] produced in the context of the IPCC requirements recognises the issue of uncertainties in emission estimates. Reliability estimates are given in conformity with the format given [79], and 'average' (i.e. expected) actual degrees of reliability were recorded down to the sub-classification level 34 .


The current emission inventory follows the IPCC guidelines for reporting to the UN Climate Convention [80], and in the context of uncertainties, is restricted to qualitative data quality indicators. Official Swedish emission statistics do not include data on emissions of N
2O, CH4 and CO. Estimates of the size of the emissions are therefore highly uncertain in many cases.

United States of America

Key sources of uncertainty that affect the projection of future emissions are listed [81], but no attempt is made to quantify these projections further, or current uncertainties in estimates. The factors themselves comprise, Climate Change Action Program funding, legislation, energy prices, economic growth, electricity supply demand, forest carbon sequestration and current climate. Most of these factors are acknowledged to be able to affect emissions either positively or negatively.

A3 Conclusions of Review of Approaches

The methods reviewed above span the entire range and provenance of information in an emission inventory. By way of a summary, Table A8 presents the features and benefits of the main types of the analytical methods, and indicates the relevance of each in the present context. This review of approaches was completed prior to the completion of the full assessment of the uncertainties in the emissions inventory. As such, it was not clear which of the reviewed methods would prove to be most useful in the subsequent analysis. Where the quality of information is most satisfactory, it was the intention to use the @RISK probabilistic tool to investigate how the uncertainties in the input parameters are transmitted through to the final inventory. Where the data quality proved to be less satisfactory, the less quantitative methods would be considered, with the semi-quantitative techniques and rating systems considered as a priority.

In the event, as noted in the Introduction of the main report, sufficient information was uncovered, or sufficiently robust judgements could be made, to justify the use throughout this study of the quantitative methods outlined in Section A1.3 of this Appendix. As this proved possible, the potentially conceptual difficulty of how to present some overall measure of uncertainty when this is contributed to by components which have varying degrees of quantitativeness did not arise. The issue of the variability of parameter values about a mean, as opposed to uncertainty around a 'true' value, however, did remain.