Novel Application of Machine Learning Techniques for Rapid SourceApportionment of Aerosol Mass Spectrometer Datasets

作者：Pande, Paritosh; Shrivastava, Manish^*; Shilling, John E.; Zelenyuk, Alla; Zhang, Qi; Chen, Qi; Ng, Nga Lee; Zhang, Yue; Takeuchi, Masayuki; Nah, Theodora; Rasool, Quazi Z.; Zhang, Yuwei; Zhao, Bin; Liu, Ying

来源：ACS Earth and Space Chemistry, 2022, 6(4): 932-942.

DOI：10.1021/acsearthspacechem.1c00344

• aerosol mass spectrometer
• SOA
• machine learning
• logistic regression
• classification
• ensemble regression
• source apportionment

We apply machine learning approaches sparse multinomial logistic regression to classify aerosol mass spectrometer(AMS) unit mass resolution (UMR) data followed by an ensemble regression technique for source apportionment of organicaerosols (OA). The classifier was trained on 60 well characterized laboratory and positive matrix factorization (PMF) deconvolvedreference spectra to identify eight OA types. These include four laboratory-derived secondary organic aerosol (SOA) spectra, whichinclude isoprene photooxidation SOA, isoprene epoxydiols (IEPOX) SOA, a monoterpene SOA type that includes alpha-pinene and beta-pinene SOA, and aromatic SOA from oxidation of naphthalene andm-xylene precursors, as well as PMF deconvolved spectra forthree primary organic aerosol (POA) types, namely, hydrocarbon-like organic aerosol (HOA), biomass burning organic aerosol(BBOA), and cooking OA (COA), and a more oxidized oxygenated OA type (MO-OOA). A 5-fold cross-validation strategy,repeated 10 times, was used to assess the classifier's performance. The classifier had high classification accuracy for COA, aromaticSOA, and isoprene SOA spectra but incorrectly classified similar to 9% by number of MO-OOA spectra as BBOA, 12% of BBOA spectra asHOA (and vice versa), and 18% of IEPOX-SOA spectra as aromatic SOA. Next, an ensemble regression model was trained on anartificially generated dataset consisting of mixtures of different OA types to assess its ability to predict fractional mass abundancesfrom classification probabilities of various OA species obtained from the multinomial logistic regression classifier trained on thereference spectra. Ultimately, the proposed approach was applied for source apportionment of aircraft-based AMS measurements ofOA UMR spectra during the HI-SCALEfield campaign. On two representative days (May 6th and 18th, 2016), the algorithmdetermined that similar to 50-60% of OA by mass was MO-OOA, which represented a highly aged organic aerosol mixture from differentsources. On both days, BBOA was determined to contribute less than 10% to OA by mass. However, on May 18th, the aromaticSOA fraction was higher compared to that on May 6th. The proposed approach is capable of rapidly analyzing AMS data in realtime, making it suitable for applications where rapid source apportionment of AMS OA spectra is desirable.

• 单位
  北京大学