Joint work with Nashat Ahmad, David Bacon, Zafer Boybeyi, Mary Hall, and Pius Lee.
Resolving the flow and pollutant concentrations at the highest possible resolution is of paramount importance in air quality and atmospheric chemistry calculations especially when dealing with chemical reactions in plumes which can vary in scales ranging from a few meters near the sources to several hundred kilometers at considerable distances from the source. The Operational Multiscale Environment model with Grid Adaptivity (OMEGA) is used to explore the modeling of pollutant plumes. OMEGA is built upon an unstructured adaptive grid made up of triangular prisms. OMEGA also has an embedded Lagrangian Atmospheric Dispersion Model (ADM). The ADM uses a puff approach to disperse the pollutants. The particles are treated as centroids of growing puffs, with the growth determined by the ambient turbulent characteristics. It also features a particle diffusion algorithm using a Monte Carlo method with a receptor- oriented concentration calculation algorithm. The basic motivation of grid adaptivity is the ability to increase resolution just in the area of interest and keep the domain coarse elsewhere to decrease computational costs. In current nested-grid models, a-priory knowledge of the solution is not required to determine where high-resolution nests are to be placed. This is not the case with OMEGA, as the grid resolution is automatically adapted to the evolving solution. The OMEGA grid adapts to the model solution via a sequence of grid refinement and coarsening functions. These are controlled by a cost function built from criteria imposed on the various model variables. Thus it is possible to adapt easily to more than one variable simultaneously. In this paper, we will discuss the OMEGA architecture, adaptation methodology, and present some examples of the dynamic grid-adaptation. In particular we will discuss an application to plume-chemistry on which the adaptivity criteria are set to dynamically adapt the grid around the predicted plume during the simulation. The cost function is built on the location of the puff centroids. This allows the plume to feelš the terrain and other environmental parameters at a higher resolution without adding high resolution elsewhere, and without any a priori knowledge of the solution. We will present results of several runs using the solution adaptive refinement and compare them with traditional fixed-resolution results.