Satellite observations of aerosols closing open cells: cloud structures and radiative forcing

Biomass burning (BB) emissions are a major global source of black carbon (BC) aerosol of which agricultural burning and deforestation fires account for at least a quarter of emissions. However, high temporal and spatial variability in the emissions lead to major challenges for their quantification and representation in large scale models. Previous modelling studies have found a persistent underestimation of aerosol optical depth (AOD) observed in BB-influenced regions, requiring particulate matter (PM) emissions from fires to be scaled by a factor of ~2-4 in order to match observations. Here we evaluate a global aerosol microphysics model (GLOMAP) against long-term observations of PM, BC and AOD to better understand tropical BB aerosol and quantify its direct radiative effects (DRE). Our region of study is Amazonia, where large scale BB is the dominant source of fine mode PM and BC. We performed simulations with GLOMAP for the period 2000-2012 using two different BB emission inventories: GFED3 and GFASv1.0. Simulated monthly mean aerosol concentrations over Amazonia are comparable between the two emission inventories, with slightly longer BB seasons and a lower peak in 2010 with the GFAS emissions. We use measurements of PM2.5 (made using gravimetric filter analysis) and equivalent BC (made using optical reflectance and Multi-angle Absorption Photometry) at two ground sites in Amazonia during 2008-2012. Comparisons with these observations show that the model captures the seasonal cycle of PM2.5 and BC mass concentrations well (R2 between 0.7 and 0.8) with maximum concentrations predicted in the dry season (~July to October). However, the model substantially underpredicts the observed concentrations of BC mass at these sites throughout the 2008-2012 period (bias GFED, -63%; GFAS -75%), despite a relatively good agreement with observed total PM2.5 (bias GFED, -15%; GFAS, -28%). Over Amazonia, the uncertainties in modelled BC mass are dominated by uncertainties in the BB emissions flux, thus the negative bias is most likely due to errors in the emissions, although microphysical processes may also play a role. Increasing GFED3 emissions of BC by a factor of 3.4 reduces the model bias in the 2008, 2009 and 2011 dry seasons but results in an overprediction in 2010 (overall bias of 14%). To further investigate the underprediction of BC mass over Amazonia, we use AOD observations from AERONET and MODIS, ground based optical measurements, and aircraft observations of BC from the South American Biomass Burning Analysis (SAMBBA) field campaign in September 2012. Using the Edwards-Slingo radiation transfer model we calculate that the 2008-mean DRE of BB aerosol over South America is negative with the standard GFED3 emissions (-0.09 W m-2), but if the BC emissions are scaled upwards to match observations the DRE becomes positive (0.12 W m-2). These results indicate that an underprediction of BC emissions from BB sources has important implications for quantifying the DRE of deforestation fires and agricultural burning.