Predicting atmospheric aerosol size distributions using Mixture Density Networks

Joshua J. Rudiger; John Stephen deGrassie; Kevin McBryde; Stephen Hammel

doi:10.1117/12.2189931

Translator Disclaimer

4 September 2015 Predicting atmospheric aerosol size distributions using Mixture Density Networks

Joshua J. Rudiger, John Stephen deGrassie, Kevin McBryde, Stephen Hammel

Author Affiliations +

Proceedings Volume 9614, Laser Communication and Propagation through the Atmosphere and Oceans IV; 96140M (2015) https://doi.org/10.1117/12.2189931
Event: SPIE Optical Engineering + Applications, 2015, San Diego, California, United States

Abstract

We use Mixture Density Networks (MDN) to estimate atmospheric particle size distributions based upon metrological parameters. Measurements of particle size spectra show that distributions are often multi-modal, composed of various underlying aerosol species that can grow from one mode to another. The flexibility of the MDN allows for the prediction of an arbitrary distribution. We show here that the MDN prediction engine can be useful in forecasting complicated, multi-modal particle size distributions. To inform and train the MDN we use meteorological, particle size and optical extinction measurements taken from a three week propagation measurement field campaign.

Citation Download Citation

Joshua J. Rudiger, John Stephen deGrassie, Kevin McBryde, and Stephen Hammel "Predicting atmospheric aerosol size distributions using Mixture Density Networks", Proc. SPIE 9614, Laser Communication and Propagation through the Atmosphere and Oceans IV, 96140M (4 September 2015); https://doi.org/10.1117/12.2189931

ACCESS THE FULL ARTICLE