Biomass burning CO, PM and fuel consumption per unit burned area estimates derived across Africa using geostationary SEVIRI fire radiative power and Sentinel-5P CO data

We present the first top-down CO fire emissions inventory for Africa based on the direct relation between geostationary satellite-based fire radiative power (FRP) observations and polar-orbiting satellite observations of total column carbon monoxide (TCCO). This work significantly extends the previous Fire Radiative Energy Emissions (FREM) approach that derived total particulate matter (TPM) emission coefficients from FRP and aerosol optical depth (AOD) observations. The use of satellite-based CO observations to derive biome-specific CO emission coefficients, ECCOb, addresses key uncertainties in the use of AOD observations to estimate fire-generated CO emissions including the requirement for a smoke mass extinction coefficient in the AOD to TPM conversion and the large variation in TPM emission factors - which are used to convert TPM emissions to CO emissions. We use the FREM-derived CO emission coefficients to produce a pan-African CO fire emission inventory spanning 2004 to 2019. Regional CO emissions are in close agreement with the most recent version of GFED(v4.1s), despite the two inventories using completely different satellite datasets and methodologies. Dry matter consumed (DMC) and DMC per unit burned area are generated from our CO emission inventory - the latter using the 20 m resolution Sentinel-2 FireCCISFD burnt area (BA) product for 2019. We carry out an evaluation of our FREM-based CO emissions by using them as input in the WRF-CMAQ chemical transport model and comparing simulated TCCO fields to independent Sentinel-5P TROPOMI TCCO observations. The results of this evaluation show FREM CO emissions to generally be in good agreement with these independent measures - particularly in the case of individual fire-generated CO plumes, where modelled in-plume CO was within 5 % of satellite observations with a coefficient of determination of 0.80. Modelled and observed total CO, aggregated over the full model domain, are within 4 % of each other, though localised regions show an overestimation of modelled CO by up to 50 %. When compared to other evaluations of current state-of-the-art fire emissions inventories, the FREM CO emission inventory derived in this work shows some of the best agreement with independent observations. Updates to previously published FREM TPM emissions coefficients based on this methodology are also provided, along with a similar evaluation as conducted for CO. The methodology described in this work is forming the basis of a forthcoming near-real-time fire emissions product from Meteosat to be issued by the EUMETSAT LSA SAF (https://landsaf.ipma.pt/en/, last access: 19 December 2022).