Application of WRF/Chem over North America under the AQMEII Phase 2: Part I. Comprehensive evaluation of 2006 simulation

摘要

The Weather Research and Forecasting model with Chemistry (WRF/Chem) version 3.4.1 has been modified to include the Carbon Bond 2005 (CB05) gas-phase mechanism, the Modal for Aerosol Dynamics for Europe (MADE) and the Volatility Basis Set (VBS) approach for secondary organic aerosol (hereafter WRF/Chem-CB05-MADE/VBS), and aerosol-cloud-radiation feedbacks to improve predictions of secondary organic aerosols (SOA) and to study meteorology-chemistry feedbacks. In this Part I paper, a comprehensive evaluation is performed for WRF/Chem-CB05-MADE/VBS to simulate air quality over a large area in North America for the full year of 2006. Operational, diagnostic, and mechanistic evaluations have been carried out for major meteorological variables, gas and aerosol species, as well as aerosol-cloud-radiation variables against surface measurements, sounding data, and satellite data on a seasonal and annual basis. The model performs well for most meteorological variables with moderate to relatively high correlation and low mean biases (MBs), but with a cold bias of 0.8-0.9 degrees C in temperature, a moderate overprediction with normalized mean biases (NMBs) of 17-22% in wind speed, and large underpredictions with NMBs of -65% to -62% in cloud optical depths and cloud condensation nuclei over the ocean. Those biases are attributed to uncertainty in physical parameterizations, incomplete treatments of hydrometeors, and inaccurate aerosol predictions. The model shows moderate underpredictions in the mixing ratios of O-3 with an annual NMB of -12.8% over rural and national park sites, which may be caused by biases in temperature and wind speed, underestimate in wildfire emissions, and underestimate in biogenic organic emissions (reflected by an NMB of -79.1% in simulated isoprene mixing ratio). The model performs well for PM2.5 concentrations with annual NMBs within +/- 10%; but with possible bias compensation for PM2.5 species concentrations. The model simulates well the domainwide organic carbon and SOA concentrations at two sites in the southeastern U.S. but it overpredicts SOA concentrations at two sites and underpredicts OC at one site in the same area. Those biases in site-specific SOA and OC predictions are attributed to underestimates in observed SOA, uncertainties in VOC emissions, inaccurate meteorology, and the inadequacies in the VBS treatment. Larger biases exist in predictions of dry and wet deposition fluxes of gas and PM species due mainly to overpredictions in their concentrations and precipitation, uncertainties in model treatments of deposition processes, and uncertainties in the CASTNET dry deposition data. Comparison of WRF and WRF/Chem simulations shows that the inclusion of chemical feedbacks to meteorology, clouds, and radiation results in improved predictions in most meteorological variables. Aerosol optical depth correlates strongly with aerosol concentration and cloud optical depth. The relationships between the aerosol and cloud variables are complex as the cloud variables are not only influenced by aerosol concentrations but by larger-scale dynamical processes. (C) 2014 Elsevier Ltd. All rights reserved.