American Association for Aerosol Research - Abstract Submission

AAAR 38th Annual Conference
October 5 - October 9, 2020

Virtual Conference

Abstract View


Estimation of the Mass Absorption Cross Section of Atmospheric Black Carbon using Regression and Machine Learning Approaches

HANYANG LI, Andrew May, The Ohio State University

     Abstract Number: 63
     Working Group: Carbonaceous Aerosol

Abstract
Mass absorption cross section of black carbon (MACBC) describes the absorptive coefficient per unit mass of black carbon and is thus an essential parameter to estimate the radiative forcing of black carbon. In the atmosphere, due to the complex physicochemical properties and different mixing states of BC, the values of MACBC have been observed to vary from 2.3 to 18 m2 g-1 at different wavelengths. Many studies have sought to estimate MACBC from a theoretical perspective, but these studies require the knowledge of aerosol complex refractive indices, mixing state, and morphology, which are difficult and/ or labor-intensive to measure. This work attempts to investigate the alternative data-driven approaches (including multivariate regressions, support vector machine, and neural networks) in estimating MACBC for aerosols from various environments.

Our model utilizes multi-wavelength absorption and scattering coefficients as well as aerosol size distribution as input variables to estimate MACBC at 870 nm. The model has been applied to two field campaigns, the Two-Column Aerosol Project (TCAP) and the Cloud, Aerosol and Complex Terrain Interactions (CACTI) project, as well as biomass burning emissions collected during the Fire Influence on Regional to Global Environments Experiment (FIREX) laboratory campaign.

We assessed the applicability of the proposed approaches in estimating MACBC using statistical metrics (e.g., coefficient of determination (R2), fractional error, and fractional bias) and visual comparisons against different aerosol properties (e.g., aerosol compositions and Absorption Ångström Exponent). Overall, the approaches used in this study can estimate MACBC appropriately, but the prediction performance varies across approaches and atmospheric environments (for example, R2 ranges from 0.45 to 0.80 for the training and test datasets, and from 0.15 to 0.50 for the independent validation datasets). Ultimately, the data-driven approaches presented in this work provide a new insight to estimate MACBC with high accuracy.