1. U.S. Department of Energy Carbon Dioxide Information Analysis Center (CDIAC) Version 2017 (Boden et al., 2017) for 1751 to 2014. Emissions included in this database are those due to fossil fuel consumption (e.g., oil, coal, and natural gas), gas flaring, and cement production. Emissions are listed by country and fuel type; bunker fuels are available separately but not included in the country totals.

2. U.S. Energy Information Administration (EIA 2017) for 1980 to 2015. CO₂ emissions from the consumption of energy, including emissions resulting from the consumption of petroleum, natural gas, and coal, as well as from natural gas flaring. Emissions are computed from consumption statistics for each fuel type by applying emissions factors. Data include nonfuel use of petroleum such as asphalt for street paving and exclude emissions from geothermal power generation, cement production and other industrial processes, or municipal solid waste combustion.

3. Fossil Fuel Data Assimilation System (FFDAS) Version 2 (Rayner et al., 2010; Asefi-Najafabady et al., 2014) for 1997 to 2012. Emissions other than power production (which use a pointwise bottom-up dataset) are estimated using data assimilation to constrain a modified Kaya identity model. The two observed fields are space-based nightlights and population density. Country totals are then created by aggregating gridded emissions using Lloyd et al. (2016, 2017) gridded country boundaries based on the Database of Global Administrative Areas, called GADM. Version 2 of FFDAS produces estimates for electricity-production, industrial, residential, commercial, and transportation (other than domestic aviation and domestic waterborne) sectors and includes a posterior uncertainty as produced by the assimilation system and prior uncertainty estimates. These map closely to the Intergovernmental Panel on Climate Change (IPCC) 1A fuel consumption category (excepting 1A3a, civil aviation, and 1A3d, navigation).

4. Emissions Database for Global Atmospheric Research (EDGAR) Version 4.3.2 (Janssens-Maenhout et al., 2017a) for 1970 to 2012. Total used of all emissions listed in “CO2_excl_short-cycle_org_C” from version 4.3.2, which includes IPCC categories (see Table E.1, this page, for a partial list).

5. Emissions Database for Global Atmospheric Research Fast Track (EDGAR FT) EDGAR Version 4.3.2 FT2016 (Janssens-Maenhout et al., 2017b; Olivier et al., 2017) for 1970 to 2016. Sectoral coverage is described as “Transport, Other Industrial Combustion, Buildings, Noncombustion, Power Industry.” For unknown reasons, EDGAR FT and the standard EDGAR emissions do not agree during their common years (i.e., 2012 and before).

The CDIAC time series was chosen to represent fossil fuel emissions from Canada, the United States, and Mexico from 2004 to 2013. In part, this is due to CDIAC’s long historical coverage for all three countries and its clear definition of what goes into the country totals (e.g., Marland et al., 2007). Assigning an uncertainty to the CDIAC time series is a challenge. Andres et al. (2014) discuss various ways to characterize the uncertainty of the CDIAC product and suggest that a time-average uncertainty for the United States could be about 4% (2 standard deviations).

SOCCR1 (Marland et al., 2007) suggests ±5% for developed countries, concordant with International Energy Agency (IEA 2005; Marland et al., 2007) intercomparisons for developed countries (also 5%). Here, the fractional range of the five inventories listed previously is used, averaged over time, to represent the uncertainty. Note that some of these differences are driven by categorical differences in what is included, or not included, in the global inventories. The CDIAC time series is recognized as different from the mean of the five inventories. Results are shown in Table E.2 and Table E.3, this page, and Figure E.1.