Biomass burning is a major source of aerosols that affect air quality and the Earth's radiation budget. Current estimates of biomass burning emissions vary markedly due to uncertainties in biomass density, combustion efficiency, emission factor, and burned area. This study explores the modeling of biomass burning emissions using satellite-derived vegetative fuel loading, fuel moisture, and burned area across Contiguous United States (CONUS). The fuel loading is developed from Moderate-Resolution Imaging Spectroradiometer (MODIS) data including land cover type, vegetation continuous field, and monthly leaf area index. The weekly fuel moisture category is retrieved from AVHRR (Advanced Very High Resolution Radiometer) Global Vegetation Index (GVIx) data for the determination of fuel combustion efficiency and emission factor. The burned area is simulated using half-hourly fire sizes obtained from the GOES (Geostationary Operational Environmental Satellites) Wildfire Automated Biomass Burning Algorithm (WF_ABBA) fire product. By integrating all these parameters, quantities of PM2.5 (particulate mass for particles with diameter <2.5 µm) aerosols are calculated for each individual fire at an interval of half hour from 2002-2005 across CONUS. The PM2.5 estimates indicate that the annual PM2.5 emissions are 3.49 x 105, 3.30 x 105, 1.80 x 105, and 2.24 x 105 tons for 2002 (April to December), 2003, 2004, and 2005, respectively. Among various ecosystems, forest fires release more than 44% of the emissions although the related burned areas only account for less than 30%. Spatially, PM2.5 emissions are larger in California for all these years, but only for some individual years in Oregon, Montana, Arkansas, Florida, Arizona, Louisiana, and Idaho. Finally, the calculated PM2.5 emissions are evaluated using national wildfire emission inventory data (NWEI) and compared with estimates from different fuel loadings. The difference between NWEI and GOES fire-based estimate is less than 20% if the same fuel data are used. The evaluation suggests that the biomass burning emissions derived from multiple satellite data provide realistic spatiotemporal patterns and can be assimilated into air quality models for forecasts in real or near real time. © 2008 Elsevier Ltd. All rights reserved.