10th International Aerosol Conference September 2 - September 7, 2018 America's Center Convention Complex St. Louis, Missouri, USA
Abstract View
Temporal Variability of Submicron Organic Aerosol PMF Factor Mass Spectra During the Houston Aerosol Characterization and Health Experiment
NANCY SANCHEZ, Henry Wallace, Benjamin Schulze, Rivkah Gardner-Frolick, James Flynn, Barry Lefer, Robert Griffin, Rice University
Abstract Number: 1063 Working Group: Source Apportionment
Abstract Factor analysis techniques such as positive matrix factorization (PMF) have been widely applied to deconvolute submicron organic aerosol (OA) concentrations observed at multiple urban and rural locations. Despite the extended use of this method, PMF-based OA source apportionment has potential limitations associated with the underlying model assumption of constant PMF factor mass spectral profiles. As OA is emitted and formed in the atmosphere, continuous chemistry (i.e., aging and atmospheric processing) and varying environmental conditions (e.g., temperature and relative humidity) are expected to impact its chemical and physical character over time. The time evolving nature of OA suggests that PMF factors with invariant mass spectral signatures reflect a time average PMF factor composition, which might not fully capture the dynamics of OA character and contributing emission sources during field studies, leading, for instance, to large model residuals during certain sampling periods.
In this study, the degree of variation of OA PMF factors mass spectral signatures and its significance as a potential limitation of PMF analysis have been evaluated by analyzing non-refractory submicron OA concentrations measured in the Houston area during the 2013-2015 Houston Aerosol Characterization and Health Experiment (HACHE). As part of the HACHE study, an Aerodyne high-resolution time-of-flight aerosol mass spectrometer was deployed in a mobile air quality laboratory, and stationary sampling was conducted at over twelve locations across the Houston area with data collection spanning between two and four weeks for each location.
Different continuous and discontinuous time segments of OA mass-resolved concentrations at distinct HACHE sampling sites were analyzed by independent PMF modeling. The O:C and H:C elemental ratios of the different factors in each time-segmented model were determined and their diurnal variation was established. Inter-comparison of the resulting PMF factor mass signatures of the different time-segmented models was conducted based on different similarity metrics including the spectral contrast angle (θ) and the mass/intensity weighted cosine. This comparison indicated significant differences between the mass spectral profile of PMF factors identified as semi-volatile oxygenated, biomass burning and cooking OA (COA) at different time intervals (θ exceeding 20° for most time intervals). Less variability was generally observed for the mass signature of PMF factors classified as low-volatility oxygenated and hydrocarbon-like OA (HOA) in the distinct time segmented models (θ usually below 10°).
Marked differences in the mass spectrum of COA at different time segments were particularly evident for the data under analysis, indicating potential large variation of the sources and dynamics of this aerosol component. Detailed analysis of the COA mass spectral signatures provided insight into the specific character of this factor at different time intervals and allowed its sub-classification as less and more oxidized COA. Although similarity metrics indicated a consistent mass spectrum for factors such as HOA, the differing abundance of specific gasoline and diesel-related oxygenated mass fragments at the different time-segmented models revealed important differences in the character of this factor, indicating that even for factors with highly similar mass spectra, PMF analysis on a time-segmented basis would allow better discerning the variability in the contributions of different sources to the observed OA concentrations at specific locations (e.g., diesel vs. gasoline vehicle emissions).
The results of this study suggest that as significant temporal variation in the mass spectral profiles of the PMF factors is likely to occur, the application of PMF to lumped OA concentration data might be insufficient to reveal the dynamic character of the different OA fractions. PMF analysis of specific time segments could overcome this limitation.