Size-resolved simulations of the aerosol inorganic composition with the new hybrid dissolution solver HyDiS-1.0 - Description, evaluation and first global modelling results.

The dissolution of semi-volatile inorganic gases such as ammonia and nitric acid into the aerosol liquid phase has an important influence on the composition, hygroscopic properties and size distribution of atmospheric aerosol particles. The representation of dissolution in global models is challenging due to inherent issues of numerical stability and computational expense. For this reason, simplified approaches are often taken, with many models treating dissolution as an equilibrium process. In this paper we describe the new dissolution solver HyDiS-1.0 that was developped for the global size-resolved simulation of aerosol inorganic composition. The solver applies a hybrid approach, which allows some particle size increments to establish instantaneous gas-particle equilibrium while others are treated time dependently (or dynamically). Numerical accuracy at a competitive computational expense is achieved by using several tailored numerical formalisms and decision criteria, such as for the time- and size-dependent choice between the equilibrium and dynamic approaches. The new hybrid solver is shown to be in good to excellent agreement with a fully dynamic solver and to have numerical stability across a wide range of numerical stiffness conditions encountered within the atmosphere. We present first results of the solver's implementation into a global aerosol microphysics and chemistry transport model. We find that (1) the new solver predicts surface concentrations of nitrate and ammonium in reasonable agreement with observations over Europe, the US and East Asia; (2) models that assume gas-particle equilibrium will not capture the partitioning of nitric acid and ammonia into Aitken mode sized particles, and thus may be missing an important pathway whereby secondary particles may grow to radiation and cloud-interacting size; and (3) the new hybrid solver's computational expense is modest, at around 10 % of total computation time in these simulations. [ABSTRACT FROM AUTHOR]