Parallelization of a mesoscale atmospheric transport-chemistry model

Abstract

The model system DRAIS-RADM describing meteorological and chemical processes in a mesoscale area is especially used to simulate the distribution of concentrations of chemical reactants in the atmosphere. The high expense in computing time and in storage needed for solving the system of nonlinear partial differential equations on which the model system is based suggests the use of parallel computers. The variables are given on a rectangular three-dimensional grid. For parallization a scalable one- or two-dimensional domain decomposition of the grid has been choosen. The implementation has been realised on the Paragon XP/S using explicit message passing and asynchronous commmunication. Measurements of the computing time for a fixed model size using different numbers of nodes are represented.