Numerical advective schemes used in air quality models—sequential and parallel implementation

Abstract

Abstract Six algorithms for solving the advection equation are compared to determine their suitability for use in photochemical air quality models. The algorithms tested are the Smolarkiewicz method, the Galerkin finite element method, the numerical method of lines, the accurate space derivative method (ASD), Bott method, and Emde method. Four algorithms for filtering the numerical noise produced when solving the advection equation are also compared. The algorithms are evaluated both on two test problems and in the CIT model. The Galerkin finite element and the ASD methods are implemented in the CIT in parallel computation. Results indicate that the ASD method, coupled with the Forester filter, produces the most accurate results. When the ASD transport solver is implemented in parallel, a speed-up of about 88 is achieved using, 256 processors. Furthermore, a new set of optimized Forester filter parameters for grid-based air quality models is determined.