APPLICATION OF A SPARSE-MATRIX, VECTORIZED GEAR-TYPE CODE IN A NEW AIR-POLLUTION MODELING SYSTEM

Air pollution models simulate numerous physical, chemical, and dynamic al processes that can be described by ordinary differential equations. Some of these processes include gas- and aqueous-phase chemistry and aerosol condensational growth. Solving these equations accurately in a model with 25,000-250,000 grid cells requires significant computation s, especially if aerosols are size-resolved. Recently, a sparse-matrix , vectorized Gear-type code (SMVGEAR) was developed that permits a com bination of fast and accurate solutions to these equations. About half the speed improvement of SMVGEAR compared to an original Gear code wa s due to sparse-matrix techniques and the other half was due to vector ization methods. Additional speedups were obtained by predicting stiff ness in, each model grid cell, reordering grid cells according to stif fness, then solving equations for stiff cells together. The modified v ersion of SMVGEAR was used in a new gas, aerosol, transport, and radia tion model (GATOR). GATOR was coupled to the mesoscale meteorological and tracer dispersion model (MMTD), and simulation results were compar ed to observations obtained during the Southern California Air Quality Study of 1987.