We apply satellite fire detection products and air quality modeling to study the contribution of fire emissions to ambient aerosol concentrations over the southeastern U.S. We find that satellite MODIS fire counts show more extensive summer burnings than suggested by the bottom-up fire inventory VISTAS in the summer of 2002. We develop a hybrid emission inventory that combines information from satellite fire counts and the bottom-up inventory by scaling the data of top-down fire count in the other months with its ratio to the bottom-up burned area data in March, the month of most prescribed burning in the Southeast in 2002. Such computed burned areas in summer are higher than the bottom up inventory in summer; the increase of fire emissions is spatially allocated over satellite observed fire pixels based on the spatial distribution of fuel loading. We show that the updated fire emission inventory leads to notably improved CMAQ model performance of OC, EC and PM2.5, in the Southeast on a regional basis, with reduced model low bias in the summer and better agreement with the observed seasonality. Our study suggests that missing fire emissions in bottom-up inventories can partially explain the underestimated concentrations of PM2.5, OC and EC in the Southeast and demonstrates that satellite fire detection can help improve our understanding of fire emissions and their impact on air quality. © 2015 Elsevier Ltd. All rights reserved.