Implementation and testing of a new aerosol module in WRF/chem

Air quality models (AQMs) need appropriate aerosol modules to simulate the formation and fate of atmospheric aerosols in both retrospective and forecasting modes. Large uncertainties exist in aerosol treatments in the AQMs. There are two existing aerosol modules in the Pacific Northwest National Laboratory (PNNL) version of the NOAA's Weather Research and Forecasting/Chemistry Model (WRF/Chem). One is the Modal Aerosol Dynamics Model for Europe (MADE) with the secondary organic aerosol model (SORGAM) (referred to as MADE/SORGAM). The other one is the Model for Simulating Aerosol Interactions and Chemistry (MOSAIC). In this study, a detailed aerosol model, the Model of Aerosol Dynamics, Reaction, Ionization and Dissolution (MADRID), has been incorporated into WRF/Chem (referred to as WRF/Chem-MADRID). MADRID differs from the previous two aerosol modules in terms of size representation used, chemical species treated, and assumptions and numerical algorithms used. MADRID is based on a sectional representation of the particle size distribution and includes a detailed treatment of aerosol dynamics and secondary organic aerosol (SOA) formation. As an initial implementation, MADRID is coupled to the PNNL's version of Carbon-Bond Mechanism (i.e., CBM-Z). A 5-day episode (August 28 to September 2) from 2000 Texas Air Quality Study (TexAQS-2000) in the southern U.S. is used to test the WRF/Chem with MADRID with a horizontal grid spacing of 12-km. Three approaches are used in MADRID to simulate gas-particle mass transfer: equilibrium, dynamic, and hybrid approaches. All of them and two different condensational algorithms (i.e., Bott and Trajectory-Grid) used in the dynamic approach will be tested in WRF/Chem. The predictions of WRF/Chem-MADRID are evaluated against observations for gas-phase species (e.g., O₃, SO₂, NO₂, NO), and PM_2.5 and its composition. The preliminary results using the equilibrium approach for gas-particle mass transfer show that the spatial distributions of concentrations of these species are consistent with the synoptic pattern. The normalized mean biases (NMBs) are 19.8% for O₃, 305% for SO₂, 24.6% for NO₂, -60.2% for NO, and 41.7% for PM_2.5. The discrepancies between simulations and observations for these species do not increase with time. Overpredictions of O₃ occurr primarily during the nighttime. This can be attributed partially to the underpredictions of NO at night, which results in a reduced titration of O₃ by NO. The predicted individual components of PM_2.5 are compared with the observations for two sites: La Porte and Williams Tower. The predicted SO₄^2- concentrations agree well with the observations (underprediected by 0.5%). Those of NO₃^-, NH₄⁺, and Cl^- are underpredicted by 0.22 (81.5%), 0.54 (47.8%), and 0.04 mg m^-3 (80.0%), respectively. Concentrations of Na⁺ and elementary carbon (EC) are overpredicted by 0.07 (36.8%) and 1.01 mg m^-3 (147.4%), respectively. The concentrations of organic matter (OM) are underpredicted by 1.7% at La Porte but overpredicted by 163.7% at Williams Tower. Differences in observed and simulated other unknown inorganic species account for 56.3% of the discrepancies of the total PM_2.5 between simulations and observations. Discrepancies between the simulation results and observations and among results with different gas-particle mass transfer approaches and different condensational algorithms will be analyzed and discussed. The gas-particle mass transfer approach and the condensational algorithm that provide the best compromise between numerical accuracy and computational efficiency will be recommended for real-time forecasting applications with WRF/Chem-MADRID.