Atmospheric aerosol particles are composed of multiple chemical species. The per-particle composition, also called “aerosol mixing state”, is of crucial importance for assessing the aerosol climate impact, since it governs the chemical reactivity of the particles, their cloud condensation nuclei activity, and their radiative properties.

Representing the aerosol mixing state in traditional aerosol models is challenging, but progress has been made recently by a number of state-of-the-art climate models, which account for the mixing state in simplified ways. However, to this date there has not been a rigorous benchmark comparison to determine the error of these simplified mixing state treatments.

This study focuses on the comparison of the modal aerosol model MATRIX and PartMC-MOSAIC. MATRIX is used to represent aerosol microphysics in the GISS-modelE global model and simulates the interaction of 16 aerosol modes, the definition of which are based on mixing state. PartMC-MOSAIC is a particle-resolved model that tracks individual particles. It simulates the per-particle composition explicitly, does not rely on any a priori assumptions about mixing state, and can hence be used as a benchmark with respect to mixing state.

For this study we developed an error measure for the comparison of MATRIX and PartMC based on the L2 norm of the differences in number size distribution. We will show the results of this comparison for a suite of test cases. These range from the simplest case where two modes only interact by coagulation, to more complex cases where several modes are involved and both coagulation and condensation of sulfuric acid act to shape the size distribution.