Abstract View
Comparison of Inversion Schemes for Retrieving Black Carbon Mixing State Distributions using CPMA-SP2 Measurements
Naseri Arash, Timothy Sipkens, Una Trivanovic, Mohsen Kazemimanesh, Olanrewaju Wasiu Bello, Allan Bertram, Steven Rogak, JASON S. OLFERT, University of Alberta
Abstract Number: 78
Working Group: Instrumentation and Methods
Abstract
Black carbon (BC) is one of the main contributors to the anthropogenic radiative forcing that causes climate change. The short life cycle of BC and its presence in complex forms (e.g., mixed with or coated by other inorganic and organic materials) results in large uncertainties in BC contribution to radiative forcing. To lower these uncertainties, robust measurements of the distribution of refractory BC mass and non-refractory (e.g. organic) coatings of atmospheric particles is required. Often, these measurements use simplified assumptions for data analysis, e.g. irregular morphology of BC is often represented as an equivalent sphere and described with an effective density. Recently, Broda et al. (2018) proposed a novel method by employing the Twomey-Markowski inversion scheme with a CPMA-SP2 system to attain the two-dimensional refractory BC mass-total particle mass distribution, indicating the distribution of the mixing state. This study improves this method by implementing a novel Bayesian approach that combines data with an exponential distance prior that naturally correlates counts for particles that are expected to share characteristics. To start, both inversion methods are tested with some fabricated data (i.e. ‘phantoms’). Results indicated that the exponential distance method can improve the accuracy of reconstructions of narrow and broad distributions by ~60% and ~20%, respectively. Both methods were then implemented on a set of experimental data generated by injecting sodium chloride solutions, using an ultrasonic atomizer, into a turbulent diffusion flame. This resulted in particles that can be composed of sodium chloride and soot. Solutions of 0.03% and 3% salt in deionized water, which produced sharp and diffuse black carbon mixing state distributions, respectively, were used to demonstrate how the exponential distance prior can improve the description of the mixing state for each case.