Comparisons of several real-time air quality forecasts of ozone, PM2.5, and their precursors with observations collected during the 2002 and 2004 New England Air Quality Studies (NEAQS)

Much like severe weather forecasts, real-time forecasts of O₃ and PM2.5 from comprehensive, Eulerian-based air quality models are constantly under pressure to improve reliability and public usefulness. The evaluation of O₃ and PM2.5 predicted by these models with observations from various monitoring networks provides one yardstick to compare forecast performance and improvement over time. However, current air quality forecast models are massive computational systems, and their limitations are known to be due to a variety of uncertainties in their basic formulation and data sources that drive the forecasts. These include uncertainties in precursor emissions, basic gas-phase and particulate-phase conversion mechanisms, and boundary layer transport processes, just to name a few. Forecast evaluations from additional data, usually collected during intensive field studies, provide the only yardstick for measuring a model's ability to characterize the photochemical, physiochemical and transport processes affecting PM2.5 and O₃ levels. This work is intended to provide a summary of findings from forecast model evaluations of two intensive field studies carried out by NOAA/OAR; NEAQS-2002 and ICARTT/NEAQS-2004. Both of these field studies are centered over the Northeastern U.S., and both have yielded an unprecedented amount of information related to atmospheric composition, as well as its vertical and spatial variability in this region. Highlights and important findings of previous and current evaluations for six to nine real-time forecast models with observations collected during these field studies are presented. The forecast models include the NOAA/FSL WRF/Chem and MM5/Chem models, the NWS/NCEP CMAQ/ETA, the Canadian CHRONOS and AURAMS models, the Baron AMS Inc. MAQSIP-RT, and the University of Iowa STEM-2K3 model.