10th International Aerosol Conference September 2 - September 7, 2018 America's Center Convention Complex St. Louis, Missouri, USA
Abstract View
Machine Learning to Predict the Global Distribution of Aerosol Mixing State Metrics
Michael Hughes, Jack Kodros, Jeffrey R. Pierce, Matthew West, NICOLE RIEMER, University of Illinois at Urbana-Champaign
Abstract Number: 936 Working Group: Aerosol Transport and Transformation
Abstract Atmospheric aerosols are evolving mixtures of chemical species, and the term "mixing state" is often used to characterize this mixture. A population in which each particle is composed of one chemical species is called externally mixed, one where all particles are the same mixture of species is internally mixed, and real populations lie in between. In global climate models (GCMs) the aerosol mixing state is represented in a highly simplified manner. For example, modal models represent all aerosols within a given mode as internally mixed, while different modes are considered externally mixed. Failing to capture the true aerosol mixing state of ambient aerosols can introduce errors in the estimates of climate-relevant aerosol properties, such as the concentration of cloud condensation nuclei.
The goal for this study is to determine a global spatial distribution of aerosol mixing state with respect to hygroscopicity, as quantified by the mixing state metric χ. In this way, areas can be identified where the external or internal mixture assumption is more appropriate. Calculating χ requires detailed per-particle composition information, which is not possible to directly predict in GCMs. We therefore used the output of a large ensemble of particle-resolved box model simulations in conjunction with machine learning techniques to train a model of the mixing state metric χ. This lower-order model for χ uses as inputs only variables known to GCMs, enabling us to create a global map of χ based on GCM data. We found that χ varied between 20% and nearly 100%, and we quantified how this depended on particle size, location, and time of the year. This framework demonstrates how machine learning can be applied to bridge the gap between detailed process modeling and a large-scale climate model.