Six Global Biomass Burning Emission Datasets: Inter-comparison and Application in one Global Aerosol Model

Abstract. Aerosols from biomass burning (BB) emissions are poorly constrained in global and regional models, resulting in a high level of uncertainty in understanding their impacts. In this study, we compared six BB aerosol emission datasets for 2008 globally as well as in 14 sub-regions. The six BB emission datasets are: (1) GFED3.1 (Global Fire Emissions Database version 3.1); (2) GFED4s (Global Fire Emissions Database version 4 with small fires); (3) FINN1.5 (Fire INventory from NCAR version 1.5); (4) GFAS1.2 (Global Fire Assimilation System version 1.2); (5) FEER1.0 (Fire Energetics and Emissions Research version 1.0), and (6) QFED2.4 (Quick Fire Emissions Dataset version 2.4). Although biomass burning emissions of aerosols from these six BB emission datasets showed similar spatial distributions, their global total emission amounts differed by a factor of 3–4, ranging from 13.76 to 51.93 Tg for organic carbon and from 1.65 to 5.54 Tg for black carbon. In most regions, QFED2.4 and FEER1.0, which are based on the satellite observations of fire radiative power (FRP) and utilize the aerosol optical depth (AOD) from the Moderate Resolution Imaging Spectroradiometer(MODIS), yielded higher BB emissions than the rest by a factor of 2–4. In comparison, the BB emission from GFED4s and GFED3.1, which are based on satellite retrieval of burned area and no AOD constraints, were at the low end of the range. In order to examine the sensitivity of model simulated AOD to the different BB emission datasets, we ingested these six BB emission datasets separately into the same global model, the NASA Goddard Earth Observing System (GEOS) model, and compared the simulated AOD with observed AOD from the AErosol RObotic NETwork ( AERONET) and MODIS in 14 sub-regions during 2008. In Southern hemisphere Africa (SHAF) and South America (SHSA), where aerosols tend to be clearly dominated by smoke in September, the simulated AOD were underestimated in all experiments. More specifically, the model-simulated AOD based on FEER1.0 and QFED2.4 were the closest to the corresponding AERONETdata, being about 73 % and 100 % of the AERONETobserved AOD at Alta-Floresta in SHSA, 49 % and 46 % at Mongu in SHAF, respectively. The simulated AOD based on the other four BB emission datasets accounted for only ~ 50 % of the AERONETAOD at Alta Floresta and ~ 20 % of at Mongu. Overall, during the biomass burning peak seasons, at most of the selected AERONETsites in each region, the AOD simulated with QFED2.4 were the highest and closest to AERONETand MODIS observations, followed closely by FEER1.0. The differences between these six BB emission datasets are attributable to the approaches and input data used to derive BB emissions, such as whether AOD from satellite observations is used as a constraint, whether the approaches to parameterize the fire activities are based on burned area, FRP, or active fire count, and which set of emission factors is chosen.