Sensitivity of WRF/Chem Predictions to Meteorological Schemes

Meteorological schemes used in 3-D atmospheric models affect meteorological and chemical predictions. The performance of these schemes may vary from episode to episode and depends on the horizontal grid spacing and time resolution used. In this study, the Weather Research and Forecasting/Chemistry Model (WRF/Chem), with a 12-km horizontal grid spacing is applied for the 28 August – 2 September 2000 Texas Air Quality Study (TexAQS-2000) episode over the Houston-Galveston area, Texas. WRF/Chem offers several options for meteorological physics, gas-phase mechanisms, and aerosol modules. WRF/Chem simulations are conducted with the RADM2 gas-phase mechanism, the MADE with the SORGAM aerosol model, two planetary boundary layer (PBL) schemes (the YSU and MYJ schemes), and two land-surface models (LSM) (the slab and the NOAH schemes). Measurements acquired from the TexAQS-2000 are used for model evaluation. The sensitivity of meteorological predictions (e.g., temperature (T), relative humidity (RH), wind speed and direction, and PBL height) and chemical predictions (e.g, ozone, PM2.5) to alternative meteorological schemes is studied. A preliminary evaluation shows that predictions agree well with observations, especially for T and RH with correlation coefficients of 0.9 and 0.7, respectively. The simulation with the slab and the YSU schemes gives better performance than that with the NOAH and the YSU schemes in terms of normalized mean bias (NMB) for RH (0.6% vs. -27.8%), wind speed (-0.5% vs. 2.6%), wind direction (6.5% vs. 8.8%), and PBL height (18.3% vs. 39.5%), but gives worse performance for T (-7.5% vs. -1.1%). Temporal analyses show that model discrepancies grow significantly on days 4-5 for meteorological predictions. Compared with the observations of ozone and PM_2.5 concentrations, the simulations with the NOAH and both the slab and the YSU schemes significantly underpredict daytime O₃ and significantly overpredict nighttime O₃, with a net overall overprediction because of the dominancy of the nighttime overpredictions; they significantly overpredict PM_2.5 at most sites during most daytime and nighttime periods. The simulation with the NOAH and the YSU schemes performs better than that with the slab and the YSU schemes for both O₃ (32.3% vs. 36.0%) and PM_2.5 (22.4% vs. 33.3%). Compared with the NOAH scheme, the slab scheme predicts much lower nighttime T on all days, lower daytime T on days 4-5, leading to slightly worse O₃ predictions at most sites on days 4-5 and worse PM_2.5 predictions at all sites on days 3-5. The simulation with the NOAH and the MYJ schemes is being conducted and the results will be contrasted with those from the NOAH and the YSU schemes. Further analysis is being conducted to identify all possible reasons for the discrepancies between the model results and observations, and among different model options.