Comparison of the EDGAR fossil fuel CO2 emissions to Vulcan - a high-resolution, bottom-up US fossil fuel emissions data product

Accurate quantification of fossil fuel CO2 (FFCO2) emissions is essential for understanding the global carbon cycle and a key ingredient in planning and verifying greenhouse gas emissions mitigation efforts. The EDGAR FFCO2 emissions data product has been widely used in research and policy to represent this component of the carbon cycle. EDGAR uses a variety of proxy techniques to downscale emissions to the 0.1 degree spatial scale for the entire globe. The Vulcan FFCO2 emissions data product, by contrast, constructs FFCO2 emissions from the bottom-up using a variety of local data including local air pollution reporting, traffic data, and stack monitoring. In order to assess differences and explore the spatial scale at which EDGAR best represents sub-national FFCO2 emissions, we present a comparison of Vulcan FFCO2 emissions to EDGAR on an annual basis for the conterminous U.S. The total FFCO2 emissions and six-economic sectors are compared at multiple spatial resolutions from 0.1 to 5.0 degrees. Figure 1: Logarithm of the differences at 0.1 degree resolution between Vulcan and EDGAR (Vulcan minus EDGAR) for all sectors and individual sector. “Log P” is for logarithm of positive difference values (Vulcan larger than EDGAR, “Log N” for logarithm of negative difference values (Vulcan smaller than EDGAR). Difference maps exhibit distinct spatial patterns (Figure 1), reflecting large differences in the underlying spatial distribution of the emission sources or downscaling proxies. For the total FFCO2 emissions, Vulcan has larger emissions over the eastern US and west coast, and smaller emissions in the mid-west, compared to EDGAR. Emission differences for individual sectors also display significant spatial patterns: Vulcan has significantly higher emissions from the residential/commercial sectors in the mid-west and northeastern US, versus smaller emissions in southeast and west. Similarly, larger Vulcan emissions are seen in the eastern US for the industrial, on-road transportation and non-road sectors versus smaller emissions in the western US. By contrast, larger EDGAR emissions are evident over much of the US for the cement sector. A Spearman correlation analysis of the total emissions between the two datasets found a correlation of 0.84 at the 0.1 degree resolution. The correlation values for industrial, residential/commercial, on-road transportation and non-road transportation are 0.76, 0.77, 0.66 and 0.56, respectively. The cement and utility sectors have poor correlation, with values of 0.02 and -0.36, respectively. These small correlation values for the cement and utility sectors are partly due to difference in the geolocation of these very specific point sources: 61% and 57%, respectively, of emissions points in one dataset are matched by a zero value in the other. However, in the cement sector, it is clear that the EDGAR emissions use a spatial proxy which is quite different from Vulcan's specific geolocated cement production facilities. Finally, it is useful to explore the level of aggregation at which the EDGAR dataset converges with the Vulcan emissions. We test progressive spatial aggregation with increments of 0.2, 0.5, 1, 2, and 5 degrees. The corresponding correlation values are 0.88, 0.91, 0.94, 0.97 and 0.98, indicating agreement at the 0.95 level at a 2 degree spatial resolution.