Three-dimensional modeling of ozone and particulate matter: Model improvement and evaluation

Accuracy and computationally-efficiency in representing secondary organic aerosol (SOA) is essential in an air quality model because SOA constitutes a sizeable fraction of fine particulate mater (PM2.5), which impacts human health, visibility, and climate. Two aerosol modules: Model of Aerosol Dynamics, Reaction, Ionization, and Dissolution 1 and 2 (MADRID 1 and 2) have been incorporated into EPA's Models-3 Community Multiscale Air Quality (CMAQ) modeling system to simulate SOA. MADRID 2 represents a detailed treatment for SOA formation, but it is computationally more expensive than MADRID 1. In this thesis work, a zero-dimensional CMAQ with MADRID 2 (CMAQ-MADRID 2) is applied to explore various methods for improving the computational efficiency of the SOA module. A combination of several speed-up methods used in MADRID 2_FAST can significantly reduce the CPU cost by 61 to 97% (speedup by factors of 2.5-30) with percentage deviations within +/-15% from the benchmark under four representative ambient conditions and conditions with typical ranges of temperatures and relative humidities.;CMAQ-MADRID 2_FAST can reduce CPU time by 47% in three-dimensional simulations compared to original CMAQ-MADRID 2, with a reasonable compromise on accuracy. This reduction is not as much as that from the MADRID 2_FAST box model simulations. The likely reasons include the high computational demand of the gas-phase mechanism calculation (about 44% of total CPU time), and the relatively lower biogenic VOC emissions over most of the domain.;CMAQ and CMAQ-MADRID 1 and 2 have similar performances for O3, and CMAQ-MADRID 1 and 2 give better performance for PM2.5 because of the higher predicted organic carbons (OC), though the significant OC overpredictions occur in the western U.S. that is likely due to the uncertainties in primary OC emissions in that region. CMAQ-MADRID 2_FAST is further applied for a seasonal episode (i.e., contiguous U.S., June, July, and August 2001) and the results show the reliable performance for O3 and PM2.5.;CMAQ-MADRID 2 and CMAQ-MADRID 2_FAST are evaluated along with CMAQ and CMAQ-MADRID 1 using the June 12-28, 1999 episode. CMAQ shows a generally good performance in simulating ozone (O3) and PM2.5. The O3-NOx-VOC chemical regimes are identified using the integrated reaction rates (IRRs) analysis. The integrated process rates (IPRs) analysis and correlation analysis show that, aerosol processes, cloud processes, dry deposition, and emissions are correlated or large contributors to the model biases for PM2.5 and its components. These results are used to guide the design of sensitivity simulations, focusing on uncertainties in the dry deposition velocities of particulate matter (PM) species and precursors, the emissions of PM precursors, and the cloud processes and gas-phase chemistry of sulfate (SO42-) formation. Adjusting the most influential processes/factors (i.e., emissions of ammonia (NH3) and sulfur dioxide (SO2), dry deposition velocity of nitric acid (HNO3), and gas-phase oxidation of SO2 by hydroxyl radical (OH)) in the final sensitivity simulation improves the overall performance of CMAQ in terms of SO42-, nitrate (NO3 -), and ammonium (NH4+).