We have developed an adaptive grid, urban-to-regional scale air quality model. The dynamic, solution-adaptive grid algorithm reduces the errors related to insufficient grid resolution by automatically refining the grid scales in regions with large changes in pollutant gradients. In other parts of the domain, the grid scales are coarsened. This results in a near-optimal use of available computational resources at all times during the simulation. The movement of the grid is controlled by a weight function formulated as the linear combination of error estimates for various pollutant species. The species to be included in the weight function calculation are application dependent and their determination for optimum performance is a current research topic. In preliminary applications involving dispersion and chemistry of puffs and plumes, the adaptive grid model was more accurate and efficient than static grid models.

An important component of the adaptive grid model is the emission processor that maps point, area and mobile sources onto the non-uniform grid cells after every grid adaptation. For efficiency, this processor uses customized intersection algorithms instead of more general algorithms of geographic information systems. A meteorological processor was also developed that can map the output of a uniform grid meteorological model at a very fine scale onto the adapting grid. For future applications, we are planning to develop a meteorological model that can operate on the same grid and run in parallel to the air quality model.

In this paper, after a brief description of the adaptive grid model, we report its recent application to an ozone episode in the Tennessee Valley during July 1995. A mesoscale model at 4-km resolution over the region provided the meteorological inputs. The emission inputs processed after each grid adaptation consisted of area and mobile sources mapped on a 4-km grid and over 9000 point sources. First, the estimated ozone and precursor levels are compared to estimates from static grid models using comparable computational resources. Then, they are compared to observations for an evaluation of the improvement provided by the adaptive grid model. Finally, means of further improvement in model performance are discussed.