Source Apportionment of Organic Aerosol Mass: Evaluating Molecular Composition-Based Methods

Thesis (Master's)--University of Washington, 2021Improvements in organic aerosol (OA) source apportionment techniques are investigated based on field measurements made in the Southeast US by a Chemical Ionization Mass Spectrometer (CIMS) equipped with a custom Filter Inlet for Gases and AEROsols (FIGAERO), as part of the Southern Oxidant and Aerosol Study (SOAS). Non-Negative Matrix Factorisation (NNMF) is applied to the particulate data in the form of both resolved thermograms and concentration timeseries. Assessments of the variance explained in the input data sets by the NNMF reconstructed approximation are used as a statistical tool for a less subjective choice of the number of factors. Linear correlation coefficients and vector phase angle are also used to produce a quantitative measure of the relative similarity between the output factors both temporally and in regards to composition. Each factor contains specific thermogram behavior (from which volatility information can be derived), unique weights for individual ions corresponding to individual molecular components of measured OA, and diurnal cycles. All three pieces of information were used to assign a specific source to each factor, ultimately showing that the dominant component of OA captured by the FIGAERO-CIMS stems from the oxidation of monoterpenes. Individual molecular components were permitted to belong to multiple and potentially all groupings of OA determined by NNMF, revealing certain factors with similar composition but remarkably different volatility and temporal trends. The median mass contribution determined from each factor produced by this factorisation routine, with no a priori information used as input, align well with those determined by an independent study of particle data during SOAS using a spectral basis set produced from several laboratory chamber experiments. The factorisation routine is shown to be more robust using resolved thermograms as input rather than the concentration timeseries. Of the seven factors given for the thermogram data, three were attributed to monoterpene-derived OA with respective extremely low, low, and semi-volatile behavior. These factors combined represent 68% of the total organic aerosol mass examined. Additionally, two factors were sourced to isoprene chemistry, one correlating with IEPOX-derived SOA, and the other likely relating to photochemistry and exhibiting relatively low volatility. The isporene-related factors accounted for 22% of OA mass. Notably absent in the factorisation of OA is a category exclusively capturing the behavior of particulate organic nitrates (PON). While this may be consistent with relatively low local concentrations of this class of particles, a separate factorisation was performed on only the PON in order to examine the volatility and temporal trends of these potentially important compounds. The added layer of volatility information and molecular level identification of OA composition provided by the FIGAERO-CIMS shows potential with the NNMF algorithm to reproduce atmospherically relevant sources from observations as well as providing framework to further identify chemical processes that lead to these categories based on volatility