Predicting Atmospheric Water-soluble Organic Mass Reversibly Partitioned to Aerosol Liquid Water in the Eastern United States

MARWA EL-SAYED, Siddharth Parida, Prashant Shekhar, Amy P. Sullivan, Christopher Hennigan, Embry-Riddle Aeronautical University

     Abstract Number: 284
     Working Group: Carbonaceous Aerosol

Abstract
Water-soluble organic matter (WSOM) formed through aqueous processes contributes substantially to total atmospheric aerosol. The impact of particle drying on fine WSOM was monitored during three consecutive summers in Baltimore, MD (2015, 2016, and 2017). Sample drying induced systematic evaporation of particle-phase WSOM in all three summers and was dependent on relative humidity (RH), WSOM concentrations, isoprene concentrations, and NO_x/isoprene ratios. Two classes of models, namely multivariate polynomial regression, and random forest (RF) machine learning were used to predict the amount of evaporated organic mass using the aforementioned parameters as model inputs. Different models corresponding to each class were fitted (trained and tested) to data from the summers of 2015 and 2016 based on the coefficient of determination (R²) and the root mean square error (RMSE). Further, model validation was performed using summer 2017 data. An RF model with 100 decision trees had the best performance (R² of 0.81) and the lowest normalized mean error. The relative feature importance for this RF model was calculated to be 0.55, 0.2, 0.15, and 0.1 for WSOM concentrations, RH levels, isoprene concentrations, and NO_x/isoprene ratios, respectively. The machine learning model was thus used to predict summertime concentrations of evaporated organics in Yorkville, Georgia, and Centerville, Alabama in 2016 and 2013, respectively. The results presented herein highlight the role of machine learning as a tool to analyze large datasets and elucidate complex atmospheric processes.