Monte Carlo analyisis of uncertainties in greenhouse gas emission inventories

1. Introduction

In response to the IPCC report ‘Good Practice Guidance and Uncertainty Management in National Greenhouse Inventories’ (IPCC, 2000) various countries carried out comprehensive (“Tier 2”) uncertainty analyses. The IPCC report established guidelines describing in detail how uncertainty analysis of National Inventory Reports (NIR) should be conducted. It provides the option to choose between two levels of analysis: Tier 1 is based on the application of simplified error propagation equations, whereas Tier 2 uses comprehensive Monte Carlo techniques at a more detailed level of aggregation. As part of a Dutch Monte Carlo-based uncertainty study (Ramirez et al., 2007) we have compared the uncertainty ranges in activity data and emission factors assumed in several European Tier 2 studies. Such an analysis does not only put the Netherlands assumptions in context (as informal quality control) but also helps to understand differences in uncertainty in total GHG-emissions of different countries. The goal of our research is twofold. Firstly, to compare the differences in uncertainty ranges, probability distribution functions (PDF) and correlations assumed in the input of Tier-2 analyses of six European countries. Secondly, to assess the influence of these differences on the resulting uncertainties of the total greenhouse gas emissions reported by those countries.

2. Method

The following countries have been included in our comparison of Tier 2 uncertainty studies: Austria, Flanders (Belgian province), Finland, the Netherlands, Norway and the United Kingdom. The reference year of the Tier 2 studies differ, since most countries do not conduct a Tier 2 uncertainty analysis annually. Furthermore the aggregation level of the analysis differs among the countries, therefore not all values are directly comparable. Since we conducted the comparison in order to put the ranges used in the Dutch TIER analysis in context, the chosen aggregation level for the comparison was the aggregation level of the TIER-1 analysis in the Dutch NIR. The Tier 2 studies included in our comparison are listed in Table 1.

Table 1: Overview of European studies used in this research

3. Uncertainty comparison of sub sectors

In this section, we focus on Sector 1 (Stationary Combustion) to exemplify the type of results found in our study. The results for all sectors can be found in Ramirez et al., (2007). Sector 1 is responsible for a large share of the greenhouse gas emissions, in the Netherlands it accounts for over 60 % of the reported emissions.

3.1 Sector 1: Stationary Combustion
Table 2 shows a comparison of the PDFs assumed for the activity data (upper table) and emission factors (lower table). Note that because various countries use different (types of) aggregation levels, not all PDFs are comparable. The level at which one can compare the PDFs is depicted in the table in a graphical way. Hence, a white background colour means that the uncertainty value is well comparable to the Dutch source categories. An orange colour indicates that the PDF is less comparable to the Dutch situation. Although uncertainty data for the Dutch 2004 NIR are available at a more detailed level than the ones shown in Table 2, the data has been aggregated to a higher level of aggregation in order facilitate comparison .

Table 2: Uncertainties in activity data, emission factors and PDF in Stationary Combustion

From the comparison, we conclude that most of the uncertainties used in the Dutch analysis for the activity data of the sector 1A1 liquids are larger than the ones reported for other European countries. This can be understood from the large underlying uncertainties in the activity data of the sub sectors 1A1b Petroleum Refining and 1A1c Manufacture of Solid Fuels (Olivier et al., 2005). The uncertainty of the sector 1A1 solids is comparable to those used in Finland and Flanders. The high (systematic) uncertainties reported by Austria for all sectors are based on a large difference (up to 10%) in the fuel statistics between two major Austrian institutions (Winiwarter et al., 2001). The uncertainty of the Dutch activity data for the sector 1A1 gases is higher than that of Finland but lower than for the other European countries.

The uncertainties of carbon dioxide (CO2) emission factors for stationary combustion are presented in Table 2. The uncertainties for the Dutch sector 1A1 liquids are significantly higher than in all other European countries. A possible explanation for this is that in the Netherlands ‘residual chemical gas’ constitutes a large part of this category, especially in the year 2004 (Olivier et al., 2005). The amount of ‘residual chemical gas’ in this sector is unknown in other countries. For the sector 1A1 gases, Norway reports a high uncertainty in the emission factor. At the moment, there is not enough information available to explain this. For the Netherlands, the uncertainties of emission factor in 1A2 for liquids and solids are slightly higher than those of other countries. This can be explained by a relatively high percentage of residual chemical gas and blast furnace/OF gas.

3.2 Correlations
One of the main differences between a TIER 1 and a Monte Carlo analysis is that correlations can be accounted for. In this study, we have looked at the correlations assumed between PDFs of activity data and emission factors within a given year by country and correlations assumed between different years (i.e. the base year and year of study). Note that not all correlations are applicable in all countries, because of differences in aggregation levels. Main results are:
• Most countries, including the Netherlands, fully correlate activity data, when it is used to calculate more than one emission. This is the case for example for number of animals, which are used both for calculating enteric fermentation and manure management.
• Emission factors are correlated if e.g. the same fuel is present in more subcategories.
• The activity data is, in most cases, not correlated between base year and end year. Exceptions are histosols in Norway, peat production areas in Finland, solid and other waste and cement production in Austria.
• The emission factors between base year and end year are fully correlated in all countries except for some situations in the UK. The exceptions in the UK are related to the level of aggregation and the reference to specific studies for e.g. methane emissions for open cast and coal storage.
• Most studies lack a full description of the correlation used and based on the information reported, it seems that correlation are not fully taken into account in most studies.

4 Conclusion

The results of the Monte Carlo analyses reported by the different countries are compared in Table 3. We conclude that the uncertainty in the total GHG emissions in the Netherlands are at a similar level as the uncertainties in Flanders, Finland and the random uncertainty reported by Austria. The uncertainties in the total GHG emissions in the United Kingdom, Finland with LUCF, Norway and Austria (including the systematic uncertainties) are much larger than the values found for the Netherlands.
 

Table 3 Comparison of uncertainties in Tier 2 analyses

The large uncertainty in the total GHG emissions in the United Kingdom stems from the very large uncertainty in the total N2O emissions, which in turn stems from uncertainties in the sub sectors Nitric Acid production (2σ: 230 %), N2O emissions from agricultural soils (341 %) and N2O emissions from wastewater handling (215 %). The large uncertainty in Austria stems from the assumed large systematic uncertainties and a larger share of non-CO2 greenhouse gas emissions. In Finland, the sector LUCF explains a large uncertainty in the total CO2 emissions. The Norwegian uncertainties for all types of gases are larger; also the share of non-CO2 greenhouse gas emissions is larger.

We conclude that major differences in the uncertainty of the total greenhouse gas emissions of the countries studied stem from the differences in magnitude of the uncertainty in the total N2O emissions, which vary between around 40 and 230 %. Also the relative share of non-CO2 gases in the total GHG emission, especially N2O is key to the explanation.

Documentation of the case

A. Ramirez, C. de Keizer, J.P. van der Sluijs, J. Olivier, L. Brandes (2008), Monte Carlo Analysis Of Uncertainties In The Netherlands Greenhouse Gas Emission Inventory For 1990-2004, Atmospheric Environment, 42 (35) 8263-8272. doi:10.1016/j.atmosenv.2008.07.059

Andrea Ramírez Ramírez, Corry de Keizer and Jeroen P. van der Sluijs, Monte Carlo Analysis of Uncertainties in the Netherlands Greenhouse Gas Emission Inventory for 1990 - 2004, report NWS-E-2006-58, Department of Science, Technology and Society, Copernicus Institute for Sustainable Development and Innovation, Utrecht University, Utrecht, The Netherlands; July 2006.

References

Baggott, S. L., L. Brown, R. Milne, et al., 2005. UK Greenhouse Gas Inventory 1990 to 2003. Annual Report for submission under the Framework Convention on Climate Change. AEA Technology, Didcot, United Kingdom.

Boogaerts, B. and S. Starckx, 2004. Uitvoeren Onzekerheidsbepalingen – Emissie-inventaris broeikasgassen van het Vlaamse Gewest. Det Norske Veritas, Antwerpen, Belgium.

IPCC, 2000. Good Practice Guidance and Uncertainty Management in National Greenhouse Gas Inventories. IPCC National Greenhouse Gas Inventories Programme, Technical Support Unit, Hayama, Japan. Monni, S., 2004. Uncertainties in the Finnish 2002 Greenhouse Gas Emission Inventory. Report nr. 5, 31 pp. VTT, Finland. Available online at: http://www.vtt.fi/inf/pdf/workingpapers/2004/W5.pdf

Monni, S. and S. Suri, 2003. Uncertainties in the Finnish 2001 Greenhouse Gas Emission Inventory. Report nr. 2209. VTT, Finland. Available online at: http://www.vtt.fi/inf/pdf/

Olivier, J. L. J. and L. J. Brandes, 2005. Uncertainty in the Netherlands' greenhouse gas emissions: Estimate of annual and trend uncertainty for Dutch sources of greenhouse gas emissions using the IPCC Tier 1 approach. 773201010/2005. Version 4. Netherlands Environmental Assessment Agency (MNP), Bilthoven, The Netherlands.

Ramirez Ramirez, C.A., C. de Keizer, J.P. van der Sluijs, 2007. Monte Carlo Analysis of Uncertainties in the Netherlands Greenhouse Gas Emission Inventory for 1990-2004. Department of Science, Technology and Society, Utrecht University, NWS-E-2006-58.

Rypdal, K. and L.-C. Zhang, 2000. Uncertainties in the Norwegian Greenhouse Gas Emission Inventory. Statistics Norway, Oslo, Norway.

Statistics Finland, 2005. Greenhouse Gas Emissions in Finland 1990-2003; National Inventory Report to the UNFCCC. Helsinki, Finland.

Winiwarter, W. and R. Orthofer, 2000. Unsicherheit der Emissionsinventur für Treibhausgase in Österreich. OEFZS--S-0072. Österreichisches Forschungszentrum Seibersdorf Ges.m.b.H., Seibersdorf, Austria.

Winiwarter, W. and K. Rypdal, 2001.Assessing the uncertainty associated with national greenhouse gas emission inventories: a case study for Austria. Atmospheric Environment, 35, 5425-5440.