Real-time air quality forecasting over the southeastern United States using WRF/Chem-MADRID: Multiple-year assessment and sensitivity studies

An air quality forecasting system is a tool for protecting public health by providing an early warning system against harmful air pollutants. In this work, the online-coupled Weather Research and Forecasting Model with Chemistry with the Model of Aerosol Dynamics, Reaction, Ionization and Dissolution (WRF/Chem-MADRID) is used to forecast ozone (O3) and fine particles (PM2.5) concentrations over the southeastern U.S. for three O3 seasons from May to September in 2009, 2010, and 2011 and three winters from December to February during 2009–2010, 2010–2011, and 2011–2012. The forecasted chemical concentrations and meteorological variables are evaluated with observations from networks data in terms of spatial distribution, temporal variation, and discrete and categorical performance statistics. The model performs well for O3 and satisfactorily for PM2.5 in terms of both discrete and categorical evaluations but larger biases exist in PM species. The model biases are due to uncertainties in meteorological predictions, emissions, boundary conditions, chemical reactions, as well as uncertainties/differences in the measurement data used for evaluation. Sensitivity simulations show that using MEGAN online biogenic emissions and satellite-derived wildfire emissions result in improved performance for PM2.5 despite a degraded performance for O3. A combination of both can reduce normalize mean bias of PM2.5 from −18.3% to −11.9%. This work identifies a need to improve the accuracy of emissions by using dynamic biogenic and fire emissions that are dependent on meteorological conditions, in addition to the needs for more accurate anthropogenic emissions for urban areas and more accurate meteorological forecasts.