Novel Source Apportionment Methodologies for Secondary PM2.5 during Extreme Wintertime Meteorological Conditions in the Western U.S.

Cam Phelan, Abiola Lawa, Kamaljeet Kaur, Jenna Krall, Kerry Kelly, Heather A. Holmes, Cesunica Ivey 

Traditional receptor-oriented source apportionment methods, such as positive matrix factorization and chemical mass balance, rely on vector-component analyses of robust observational records or pre-determined source fingerprints, respectively. These methods may struggle to resolve sources that are similar in composition due to source fingerprint collinearity. It is important to isolate contributions with high source specificity to guide mitigation during extreme pollution events. This work explores source apportionment techniques that permit the estimation of contributions from 19 sources during January 2016 in 11 western U.S. cities during cold air pool events. Two apportionment techniques were employed within the Community Multiscale Air Quality model to quantify contributions to 23 speciated PM2.5 components: decoupled direct method (DDM) and integrated source apportionment model (ISAM). Source contributions from both DDM and ISAM
were used to numerically optimize existing source fingerprints for subsequent use in receptor models, with the goal of using observational nudging to improve location-specific congruence while improving estimations of secondary PM2.5. Optimization increased the representation of secondary PM2.5 mass by 5.62 to 20.1% on average. Further, when considering California-specific
fuel composition for gasoline and diesel combustion fingerprints, diesel profiles exhibited the largest changes after numerical optimization. These analyses highlight the need for localized source fingerprints for apportionment, especially for analyses during extreme air pollution episodes.