What is the aerosol state in the ambient atmosphere? Or: Multiscale modeling to tackle structural uncertainty in aerosol models

The diversity in particle composition within the atmospheric aerosol is well-documented in field observations, but usually grossly oversimplified in chemistry transport models or earth system model. This is for good reasons--- to save computational cost---but comes with the trade-off of introducing considerable and difficult-to-quantify structural uncertainty in our predictions of aerosol composition, and by extension, of aerosol interactions with clouds and radiation. This presentation will illustrate how targeted particle-resolved simulations can be used to quantify structural uncertainty in more approximate aerosol models (i.e., sectional or modal models). The particle-resolved approach resolves the aerosol using individual computational particles that evolve in size and composition during their simulated lifetime in the atmosphere. I will present our model development of WRF-PartMC, a stochastic particle-resolved model embedded into the Weather Research and Forecasting Model (WRF) for explicit simulation of aerosol mixing state on the regional scale. The novel computational methods developed for this purpose include a particle-resolved emission inventory and stochastic transport algorithms. With its fully-resolved aerosol mixing state representation, WRF-PartMC allows for direct inter-model comparisons with traditional aerosol schemes used in regional and climate models. We used this modeling approach to quantify the extent to which simplifying the diversity of aerosol composition introduces errors in our estimates of cloud condensation nuclei concentration and aerosol optical properties. I’ll conclude the presentation by demonstrating how machine learning can leverage particle-resolved simulation data to efficiently bridge to the global scale.