Investigation of mixing in a coupled air quality and weather prediction model

The accuracy of operational air quality modeling is highly dependent on the accuracy of mesoscale meteorological models in simulating turbulent mixing, which can be partitioned into grid resolvable and subgrid components. Calculation of subgrid mixing is based on the assumption that there is a clear scale separation between the horizontal and vertical mixing and that the mixing is vertically dominant. In theory, this assumption is valid only when the horizontal grid scale is much greater than the scale of the PBL depth. As the horizontal grid spacing approaches 1 km or less, an emerging argument is that a fully 3-D subgrid mixing parameterization should be used. The horizontal grid spacing in air quality models used for State Implementation Plans in the USA is typically 4 km. In such a model, the atmospheric flow and pollutant concentration fields are assumed to be spatially homogeneous within a grid cell. In reality, however, temporal and spatial scales of physical and chemical processes in a grid box of 4 km horizontal spacing vary over a wide range. In particular, the photochemical reaction time scales typically cover a range of more than ten orders of magnitude. As a result, some reactions may occur so quickly that physical transport cannot quickly smooth out the local variation of the species concentrations within the grid, and non-homogeneity of the species concentrations will result within the grid box. Therefore, it is important to know how much error the disparity between the scale of turbulent mixing and that of chemical reaction in a grid box of 4 km horizontal spacing will cause in air quality modeling using this resolution. To this end, vertical distributions of pollutants from three WRF-Chem model simulations of 4 km, 1 km and 200 m horizontal resolutions are used to estimate the errors in the 4-km simulation associated with the scale disparity.