We used the NOAA-EPA Community Multiscale Air Quality (CMAQ) model to demonstrate the impact of assimilating hourly GOES Aerosol Optical Depths (AODs) on particulate matter (PM2.5, particles smaller than 2.5 µm in median diameter) predictions. Cressman assimilation scheme was used to perform the model analysis and improve the initial conditions of aerosol species in the CMAQ model. We simulated two case studies, one an east coast sulfate haze episode and the other a biomass burning event in the western United States (U.S.). Analysis of the assimilation runs showed that the aerosol assimilation improved surface PM2.5 predictions for the sulfate episode (31% under prediction with aerosol assimilation compared to 57% without aerosol assimilation) but had a negative impact for the biomass burning event. For this event, PM2.5 predictions were under-predicted by 56% without aerosol assimilation; with assimilation, the predictions were over-predicted by 50%. The impact of aerosol assimilation on surface PM2.5 predictions depends on aerosol vertical profile, aerosol type, planetary boundary layer dynamics in the model, and the quality of the satellite data. Comparison of vertical profiles of aerosol extinction in the CMAQ with CALIPSO extinction profile shows that aerosol assimilation positively impacts the surface PM2.5 predictions when aerosols are well distributed in the boundary layer as was observed for the sulfate event whereas the impact is negative when aerosols are aloft as was found for the biomass burning episode. Analysis of MODIS fine mode/coarse mode fraction shows that the coarse mode dominates for the biomass burning events and assimilation of satellite-derived AOD without scaling for the fine mode fraction contributed to over-prediction of surface PM2.5 predictions for the biomass burning episode. Time series analysis of observed vs. predicted surface PM2.5 predictions has shown that the accuracy of the boundary layer depth in the model is critical for the aerosol assimilation to have a positive impact. We are improving the assimilation methodology from a simple Cressman analysis to Gridpoint Statistical Interpolation (GSI) data assimilation system to be consistent with the efforts underway at the National Weather Service (NWS) to enhance CMAQ PM2.5 predictions using satellite data assimilation. We are developing a regional air quality data assimilation using GOES AODs using GSI for operational applications at the NWS. Preliminary results of assimilation runs using the new GSI will also be presented.