Intercomparison of black carbon emissions from biomass burning using FLEXPART-WRF and ground-based observation in Alaska

Black carbon aerosols (BC) are emitted into the atmosphere by incomplete combustion processes of fossil fuels and biomass. Especially in the Arctic region, anthropogenic emissions from mid-latitudes (e.g., China) are transported by large-scale atmospheric circulation, and local emissions such as forest fires in boreal forests and gas flares are also considered to contribute significantly. In this study, we report on the result of intercomparison for Poker Flat, Alaska, especially focusing on the biomass burning emission inventories, which still show large differences among the inventories.

Since April 2016, observations of BC and CO have been conducted at the Poker Flat Research Range (PFRR; 65.12°N, 147.49°W) in cooperation with the University of Alaska Fairbanks. The pathways of air parcels that were observed at PFRR were estimated using the Lagrangian particle diffusion model FLEXPART-WRF version 3.3. Backward calculations were performed for 20 days using 40,000 particles every 6 hours from April 2016 to December 2020. The meteorological field was calculated by a regional meteorological model (WRF) covering the Northern Hemisphere. The concentration and source attribution has been estimated using the residence time estimated by FLEXPART-WRF and emissions at each grid. ECLIPSEv6 and 6 different inventories (FINNv1,5 FINNv2.5(MODIS, MODIS+VIIRS), GFEDv4.1a, GFASv1.2, QFEDv2.5r1, FEERv1.0-G1.2) are used as the anthropogenic and biomass burning emissions, respectively.

The concentration in the wintertime was generally well reproduced by the model, and it was estimated that anthropogenic emissions in Alaska (especially domestic and transport sectors in ECLIPSEv6) were dominant in that period. It was also found that there were very large differences in the contribution of biomass burning among inventories, especially in summer when the forest fires are active. Among them, GFEDv4.1 generally succeeded in capturing large fire events, especially in 2017 and 2019 (r=0.93). FINN inventories (version 1.5, version 2.5 with MODIS, and MODIS+VIIRS) tended to underestimate such eventual increases. In contrast, QFED sometimes overestimated concentrations at large events. If we assume the ‘event period’ as observed BC concentration exceeds the 95 percentiles for the whole period, the contribution of biomass burning was estimated to be higher at the event period (47%) than that during the non-event period (22%) in the simulation with GFEDv4.1.