CHE_EDGAR-ECMWF_2015

The new CHE_EDGAR-ECMWF_2015 dataset with anthropogenic fossil CO2 emissions and their uncertainties and with a new 7×7 covariance matrix for the atmospheric transport model was compiled and tested. The fossil CO2 emissions include all long-cycle carbon emissions from human activities, such as fossil fuel combustion, industrial processes (e.g. cement) and products use, but excludes emissions from land-use change and forestry. Human CO2 emission inventories were processed into gridded maps to provide an estimate of prior CO2 emissions, aggregated in 7 main emissions groups: 1) energy production super-emitters, 2) energy production standard-emitters, 3) manufacturing, 4) settlements, 5) aviation, 6) other transport at ground level and 7) others, with estimation of their uncertainty and covariance. For the first implementation it is assumed that each emission group is fully correlated with itself and fully uncorrelated with any other group (only diagonal values are non-zero and equal to log-normal variance).

The CHE_EDGAR-ECMWF_2015 represents the 2015 global fossil CO2 emissions prior at 0.1º×0.1º resolution that has been for the first time to our knowledge bridging the inventory community and the atmospheric modelling community. In fact, the uncertainty calculations fully respect the detailed error propagation approach recommended by IPCC (2006) guidelines for GHG inventories while these datasets as prior input were processed such that the uncertainty information could be fully taken up by the ECMWF model IFS. Estimation of emission uncertainties is purely based on IPCC (2006) and IPCC-TFI (2019) emission factor and activity data uncertainty values and assumptions – mainly that emissions are fully uncorrelated. Uncertainties related to the spatial distribution (representativeness of the proxy data and their uncertainty) were not assessed in this study, but they can be included by the user on top of the calculated emission uncertainties. All calculations, performed for the year 2015, are documented so that the methodology and algorithms used can be easily adapted for any other year. The dataset can be directly used in inverse modelling, and ensemble data assimilation applications, such as those envisaged within the Copernicus Atmosphere Monitoring Service (CAMS) system.

CHE_EDGAR-ECMWF_2015 consists of 11 global NetCDF files with gridded yearly and monthly upper and lower bounds of uncertainties in % and kg·m-2·s-1 per each ECMWF group and their sum, and 1 Excel file with 16 spreadsheets with the same information listed per country (metadata, emissions, uncertainties, statistical parameters).

Calculated emissions and uncertainties of fossil CO2 have been compared to other data sets based on the country-specific data reported to UNFCCC and on fuel-specific data reported in the energy statistics of IEA. The global values and their uncertainty at a 2σ range for the CHE_EDGAR-ECMWF_2015 dataset show the lowest value of -4.7/+9.6 % or ±7.1 % range due to the methodology used. At country level the CHE_EDGAR-ECMWF_2015 dataset provides generally larger uncertainty ranges, that are reduced when more detailed information is available to reduce the uncertainties; in summary, using the information that is uniformly available for all countries a coherent uncertainty representation is obtained.

The CHE_EDGAR-ECMWF_2015 dataset has been tested to provide the ECMWF Earth system ensemble spread to characterise the CO2 atmospheric concentrations’ uncertainties in the prototype of the Copernicus CO2 Monitoring and Verification Support Capacity. Annual and monthly uncertainties have been evaluated in the ECMWF’s atmospheric transport model IFS ensemble simulations as well as the sensitivity to the spatial distribution of anthropogenic CO2 emissions (McNorton et al., 2020). Results show to be rather sensitive to the spatial distribution proxies, and most updated proxies and prior uncertainties are better adapted for data assimilation applications. This needs to be studied in a future research project, the Prototype system for a Copernicus CO2 service (CoCO2), that follows the current CHE research project.

Contribution of representativeness errors to uncertainties and time correlation are neglected in CHE_EDGAR-ECMWF_2015 and will need to be assessed in successive future studies. The estimation of global gridded emissions with their spatially and temporally distributed uncertainties constitute the backbone for atmospheric inversions to estimate anthropogenic emissions from atmospheric concentrations (Pinty et al., 2017). Dedicated satellite missions (e.g. Copernicus anthropogenic CO2 monitoring mission CO2M described in Janssens-Maenhout et al. (2020)) are being planned to monitor anthropogenic emissions from space and substantially reduce emission uncertainties. The developments in the emission uncertainty based on prior knowledge computation presented in this paper is an important preparatory step for an ensemble-based CO2 Monitoring and Verification System prototype, such as the one developed within the CHE project.