Emission ensemble approach to improve the development of multi-scale emission inventories (Preprint)

In this work, an ensemble inventory (median) is created with the aim of monitoring the status and progress made with the development of Europe-wide inventories. This ensemble inventory also allows comparing a large number of inventories at the same time, foster interactions among emission inventory developers and allow for comparing additional inventories (e.g. bottom-up ones) with all ensemble components. In contrast with other fields of applications (e.g. air quality forecast), this emission ensemble is not necessarily better than any of its components. Although it is not the more accurate inventory, it serves here as a common benchmark for the screening. We focus on differences in terms of country totals, country sectorial share and share of the country emissions to the urban areas for emissions of NOx, PM2.5, PM coarse, NMVOC, SOx and NH3. Because the emission “truth” is unknown, the approach does not tell which inventory is the closest to reality. The methodology rather screens differences between inventories, excludes differences that are not relevant and identifies among the remaining ones, those that are larger than a given threshold, and need special attention. The underlying concept is that above this threshold, differences are so large that one or both inventories must be checked.
The analysis of the ensemble and the comparison with its individual components highlight a large number of inconsistencies. While two of the three inventories behave more closely to each other (CAMS-REG and EMEP), they yet show inconsistencies in terms of the spatial distribution of emissions. These differences mostly occur for SO2, PM and NMVOC, for the industrial and residential sectors, and reach a factor 10 in some instances. Necessary improvements have been identified, in particular with EDGAR with the PM emissions from the small-scale combustion sector and SO2 from the industry and power plant sectors. The comparison with the local inventory for Poland leads to identifying another type of inconsistencies, associated to the sectorial share at country level. This is explained by the fact that some emission sources are omitted in the local inventory due to the lacking of appropriate geographically allocated activity data. The screening process led to identify some sectors and pollutants for which discussion between local and EU-wide emission compilers would be needed in order to reduce the magnitude of the observed differences (e.g. in the residential and industrial sectors). The settings used in this work (e.g. the choice of 150 urban areas or the way sectors are aggregated) are arbitrarily fixed and can easily be adapted for the purpose of other comparisons.