Uncertainties in estimating biomass burning emissions for Africa: implications for atmospheric modelling

Africa is the biggest continental source of biomass burning emissions. The emissions result in regional to transcontinental air pollution. Atmospheric model studies that address the linkage between fires and air quality have to cope with substantial uncertainties in fire emission inventories. All contemporary fire emission inventories build upon satellite information to quantify the spatial and temporal occurrence of fires, but they use different satellite sensors, detection algorithms, and estimation methods. Large discrepancies between inventories in terms of emission totals and spatial and temporal patterns are the consequence.

Most satellite products employed for fire emission estimations across Africa cannot resolve small fires, and the omission of thereof strongly lowers the accuracy of the estimated emission fluxes. The ESA Fire_cci project has recently released the first burned area product for Africa that builds upon high resolution optical imagery from Sentinel-2. By resolving small fires, it detects 60 to 110% more burned area in 2016 than estimated by the widely used, MODIS-based products GFED4s, GFED4, MCD64A1 Collection 6, FireCCI51 and FINN.

We inter-compare biomass burning emission inventories computed from these products and analyse potential sources of discrepancies. Sensitivity simulations with the WRF-Chem atmospheric chemistry model using the inventories as boundary condition complement the analysis. Modelled concentrations of atmospheric trace species are evaluated against a set of satellite observations that act as top-down constraint on the fire emission estimates.