On the efficient treatment of vertical mixing and chemistry in air pollution modelling

Reaction-Advection-Diffusion problems occurring in Air Pollution Models have very stiff reaction terms and moderately stiff vertical mixing (diffusion and cloud transport) terms which both should be integrated implicitly in time. Standard implicit time stepping is by far too expensive for this type of problems whereas widely used splitting techniques may lead to unacceptably large errors for fast species. Recently proposed alternatives are the so-called source splitting and the ros2 method combined with approximate matrix factorization (amf). In this paper we compare these two methods with the standard Strang splitting method in a framework of an operational global dispersion model. The results of the tests favour the ros2-amf scheme