PHOTOCHEMICAL NUMERICS FOR GLOBAL-SCALE MODELING - FIDELITY AND GCM TESTING

Atmospheric photochemistry lies at the heart of global-scale pollution problems, but it is a nonlinear system embedded in nonlinear transpor t and so must be modeled in three dimensions. Total earth grids are ma ssive and kinetics require dozens of interacting tracers, taxing super computers to their limits in global calculations. A matrix-free and no niterative family scheme is described that permits chemical step sizes an order of magnitude or more larger than time constants for molecula r groupings, in the 1-h range used for transport. Families are partiti oned through linearized implicit integrations that produce stabilizing species concentrations for a mass-conserving forward solver. The kine tics are also parallelized by moving geographic loops innermost and ch anges in the continuity equations are automated through list reading. The combination of speed, parallelization, and automation renders the programs naturally modular. Accuracy lies within 1% for all species in week-long fidelity tests. A 50-species, 150-reaction stratospheric mo dule tested in a spectral GCM benchmarks at 10 min CPU time per day an d agrees with lower-dimensionality simulations. Tropospheric nonmethan e hydrocarbon chemistry will soon be added, and inherently three-dimen sional phenomena will be investigated both decoupled from dynamics and in a complete chemical GCM.