Multiscale air quality modeling: Application to southern California

Abstract

A multiscale air quality model has been developed to follow accurately and efficiently the long‐range transport of pollutants emitted from urban areas. The model employs a two‐dimensional finite element scheme to follow the horizontal transport. The multiscale capability is obtained by using local finite element refinements. The model is applied to a 3‐day intensive measuring period (August 27–29, 1987) over southern California. Uniform and nonuniform grid systems are employed in the simulations. Uniform grid systems used resolutions of 5×5 km2, 10×10 km2, and 20×20 km2, horizontally. Two nonuniform grid systems are used. The first combines the 5×5 km2 grid over urban Los Angeles with the 10×10 km2 grid over the rest of the domain, while the second uses the 5×5 km2 grid followed by the 10×10 km2 grid and 20×20 km2 grid over the ocean and sparsely populated areas. Uniform coarsening of the grid impacted the urban areas by diluting NOx emissions, leading to higher levels of urban ozone predictions. The impact on regional areas was complex. The ozone being transported downwind was not followed as accurately as on the fine grid and the regional levels usually decreased. On the other hand, the artificial dilution of NOx emissions had the adverse effect of increasing regional ozone levels. Therefore grid coarsening sometimes leads to what appears to be satisfactory levels of regional ozone for the wrong reason. By using fine grids over the source areas, more satisfactory predictions of both urban and regional ozone levels were obtained. This is also computationally more efficient than using uniform fine grids over the entire domain.