Global Near-Real-Time Estimates of Biomass Burning Emissions using Satellite Active Fire Detections

We present a new technique for generating daily global estimates of biomass burning emissions suitable for use in models forecasting atmospheric chemical composition and air quality. We combine ecosystem-dependent carbon fuel databases, fire weather severity estimates, and near-real-time satellite fire detections from the MODIS instruments to estimate the amount of carbon released from active fires. Emissions of CO, NOx, and hydrocarbons are then estimated using ecosystem-dependent emission ratios. These emissions estimates have been used to provide global chemical and regional aerosol forecasts for much of 2006 using the NASA/University of Wisconsin Realtime Air Quality Modeling System (RAQMS). The largest overall uncertainty in this approach lies in inferring area burned from instantaneous active fire detections. Here we evaluate emissions for spring and summer of 2006 by intercomparing our emissions estimates with three other approaches using satellite fire detections over regional to global domains: 1) NOAA GOES fire detections over the continental US, 2) MODIS fire and landcover products over North and Central America, and 3) 8-day composite MODIS detections applied to global GFEDv2 emissions. Overall we find there are large differences in area-burned estimates, particularly for small fires. We compare RAQMS global CO predictions with observations from the MOPITT and TES instruments over regions where biomass burning is significant. CO emission estimates from burning are consistent with satellite observations over the US but we find large discrepancy in three out of four regions of large tropical burning.