Numerical experiments on MM5–CMAQ sensitivity to various PBL schemes

The Community Multiscale Air Quality (CMAQ) modeling system developed by the United States Environmental Protection Agency (USEPA) was used along with the PSU/NCAR mesoscale modeling system MM5 to study air quality modeling sensitivity to various planetary boundary layer (PBL) schemes implemented in MM5. Two 37-day periods, representing typical summer and winter periods, were selected for the simulations. The MM5 domain covered the continental US while the CMAQ domain covered much of the central and eastern US. There were 34 vertical layers in MM5 and 18 in CMAQ. The horizontal resolution was 36 km for both. MM5 was run with five of the seven commonly used PBL schemes to generate a series of gridded meteorological fields. The data were then processed and converted by the meteorology–chemistry interface processor (MCIP) to produce meteorological input for CMAQ, in conjunction with the emissions data prepared by the Sparse Matrix Operator Kernel Emissions (SMOKE) System.The sensitivity of MM5–CMAQ modeling to the various PBL schemes was assessed and quantified by comparing model output against observations from the meteorological and the air quality monitoring networks within the CMAQ domain. The meteorological variables evaluated included surface (2-m) and 850-mb temperature, 10-meter and 850 mb wind speed and direction, 2-m mixing ratio, surface pressure, and cloud fraction over 50 surface sites and 21 upper air sounding sites. The CMAQ variables included gaseous species O3, NO2, NOx, SO2 and fine particulate matter PM2.5 for over 2000 monitoring stations. A new 3-2-1 objective evaluation approach was developed to enhance the model sensitivity analysis. Results of evaluation of these variables indicate that domain-wide the performance of neither model was sensitive to the PBL schemes. Although large differences of PBL height existed across the meteorological simulations, they did not appreciably affect the CMAQ performance. On an urban scale, differences across the CMAQ simulations were large enough for the model to be considered sensitive, although no favorable PBL scheme was identified. It is suggested that domain-wide the lack of sensitivity of the MM5–CMAQ modeling system to the PBL schemes was mainly attributed to the parameterization of vertical eddy diffusivity in CMAQ and the four-dimensional data assimilation (FDDA) in MM5.