Multi-Year Application and Evaluation of WRF/Chem-ROMS over Continental U.S. from 2009 to 2012

The exchange of heat and moisture between the atmosphere and ocean affects climate, which in turn influences the spatial and vertical distributions of air pollutants and associated public health, particularly in densely-populated coastal areas. A 3-D air quality model coupled with a regional oceanic model can better simulate air-sea interaction processes and thus improve the model’s ability in reproducing the concentrations and spatial distributions of air pollutants in coastal regions. One such a model is a newly-developed online-coupled Weather Research and Forecasting model with Chemistry coupled with the Regional Ocean Modeling Systems (WRF/Chem-ROMS) by North Carolina State University. In this work, WRF/Chem-ROMS is applied at a horizontal grid resolution of 36-km over continental U.S. (CONUS) for four years (2009-2012).  The objectives are to identify likely causes for model biases for potential model improvement through a comprehensive performance evaluation and to examine the impacts of air-sea interactions on model predictions through multi-year simulations. A comprehensive model evaluation of simulated meteorology and air quality is performed using observations from available surface networks (e.g., AIRS-AQS, CASTNET, IMRPOVE, STN, and NCDC) and satellites (CERES, MODIS, OMI, and MOPITT) in terms of statistics, spatial distribution, and temporal variation.

The model evaluation of 2009 to 2012 simulations shows an overall good performance in terms of annual mean predictions for most meteorological variables such as temperature at 2-m (T2) and wind speed at 10-m (WSP10), radiative variables such as shortwave and longwave radiation, and cloud variables such as cloud fraction (CF) and cloud droplet number concentration. For example, the annual mean biases (MBs) for T2 and WSP10 are -0.7 °C to -0.1 °C and 0 to 1.1 m s^-1, respectively. The annual mean normalized mean biases (NMBs) for CF are 0.6-0.8%. The model also shows a good performance for maximum 8-h ozone (O₃) with NMBs of -22.8% to -6.0% and for fine particulate matter (PM_2.5) with NMBs of 4.5-22.9%. However, moderate-to-large wet biases are found for precipitation with NMBs of 16.7-64.6%, which is mainly caused by to the limitations of the Grell 3D cumulus parameterization used. The cloud optical thickness (COT) is also largely underpredicted with annual mean NMBs of -51.2% to -45.6% due to poor performance of some cloud variables and limitations in COT calculations. In addition, large biases are found for PM_2.5 composition such as sulfate and organic carbon and for tropospheric ozone residual. These large biases may be due to inaccurate meteorological predictions (e.g., precipitation), emissions (e.g., primary PM species), and boundary conditions for some species (e.g., O₃).  These results will provide valuable information to further enhance the skills of WRF/Chem-ROMS for the coupled atmosphere-ocean modeling through improving the model inputs, configurations, and representations of identified physical and chemical processes.