Files
Abstract
Aerosols in the troposphere directly impact the radiative transfer in Earth's atmosphere through the absorption and scattering of incoming solar radiation. Their size regime makes them particularly strong at absorbing light as their sizes are similar to the wavelengths of visible radiation, the most prominent region of the solar spectrum in the troposphere. Their heterogeneity in size, morphology and chemical composition makes their optical properties difficult to characterize, though, and their widespread distribution in the atmosphere makes them difficult to represent in climate models. In the first portion of this dissertation, I characterize the optical properties of biomass burning aerosols (BBA) from simulations of wildland fires using fuel beds that represent eco-regions present in the Southeast United States. This study demonstrates that with low fuel moisture content (representative of wildfires), the BBA optical properties depend on the eco-region of the fuel beds, while with high fuel moisture content (representative of prescribed fires), the BBA optical properties are similar for all three eco-regions. In the second portion of this dissertation, I employ machine learning methods with various aerosol data sets. In particular, I present a data-driven method for finding clusters of arbitrary shape in a large dataset of aerosol optical properties collected from ambient measurements. This novel method is based on the density of the data in a three-dimensional space and succeeds at separating different classes of light absorbing aerosol. Together, this work helps to constrain aerosol impacts on radiative transfer in the atmosphere through a combination of experimental measurements and applying data-driven techniques to better characterize ambient aerosols.