TY - JOUR
T1 - A robust clustering algorithm for analysis of composition-dependent organic aerosol thermal desorption measurements
AU - Li, Ziyue
AU - D'Ambro, Emma L.
AU - Schobesberger, Siegfried
AU - Gaston, Cassandra J.
AU - Lopez-Hilfiker, Felipe D.
AU - Liu, Jiumeng
AU - Shilling, John E.
AU - Thornton, Joel A.
AU - Cappa, Christopher D.
N1 - Funding Information:
Financial support. This research has been supported by the Na-
Funding Information:
Acknowledgements. The SOAFFEE campaign was done at Pacific Northwest National Laboratory, supported by the U.S. Department of Energy (DOE) Office of Science, Office of Biological and Environmental Research, as part of the Atmospheric Systems Research (ASR) program. PNNL is operated for the DOE by the Battelle Memorial Institute under contract DE-AC05-76RL01830.
PY - 2020/3/2
Y1 - 2020/3/2
N2 - One of the challenges of understanding atmospheric organic aerosol (OA) particles stems from its complex composition. Mass spectrometry is commonly used to characterize the compositional variability of OA. Clustering of a mass spectral dataset helps identify components that exhibit similar behavior or have similar properties, facilitating understanding of sources and processes that govern compositional variability. Here, we developed an algorithm for clustering mass spectra, the noise-sorted scanning clustering (NSSC), appropriate for application to thermal desorption measurements of collected OA particles from the Filter Inlet for Gases and AEROsols coupled to a chemical ionization mass spectrometer (FIGAERO-CIMS). NSSC, which extends the common density-based special clustering of applications with noise (DBSCAN) algorithm, provides a robust, reproducible analysis of the FIGAERO temperature-dependent mass spectral data. The NSSC allows for the determination of thermal profiles for compositionally distinct clusters of mass spectra, increasing the accessibility and enhancing the interpretation of FIGAERO data. Applications of NSSC to several laboratory biogenic secondary organic aerosol (BSOA) systems demonstrate the ability of NSSC to distinguish different types of thermal behaviors for the components comprising the particles along with the relative mass contributions and chemical properties (e.g., average molecular formula) of each mass spectral cluster. For each of the systems examined, more than 80% of the total mass is clustered into 9-13 mass spectral clusters. Comparison of the average thermograms of the mass spectral clusters between systems indicates some commonality in terms of the thermal properties of different BSOA, although with some system-specific behavior. Application of NSSC to sets of experiments in which one experimental parameter, such as the concentration of NO, is varied demonstrates the potential for mass spectral clustering to elucidate the chemical factors that drive changes in the thermal properties of OA particles. Further quantitative interpretation of the thermograms of the mass spectral clusters will allow for a more comprehensive understanding of the thermochemical properties of OA particles.
AB - One of the challenges of understanding atmospheric organic aerosol (OA) particles stems from its complex composition. Mass spectrometry is commonly used to characterize the compositional variability of OA. Clustering of a mass spectral dataset helps identify components that exhibit similar behavior or have similar properties, facilitating understanding of sources and processes that govern compositional variability. Here, we developed an algorithm for clustering mass spectra, the noise-sorted scanning clustering (NSSC), appropriate for application to thermal desorption measurements of collected OA particles from the Filter Inlet for Gases and AEROsols coupled to a chemical ionization mass spectrometer (FIGAERO-CIMS). NSSC, which extends the common density-based special clustering of applications with noise (DBSCAN) algorithm, provides a robust, reproducible analysis of the FIGAERO temperature-dependent mass spectral data. The NSSC allows for the determination of thermal profiles for compositionally distinct clusters of mass spectra, increasing the accessibility and enhancing the interpretation of FIGAERO data. Applications of NSSC to several laboratory biogenic secondary organic aerosol (BSOA) systems demonstrate the ability of NSSC to distinguish different types of thermal behaviors for the components comprising the particles along with the relative mass contributions and chemical properties (e.g., average molecular formula) of each mass spectral cluster. For each of the systems examined, more than 80% of the total mass is clustered into 9-13 mass spectral clusters. Comparison of the average thermograms of the mass spectral clusters between systems indicates some commonality in terms of the thermal properties of different BSOA, although with some system-specific behavior. Application of NSSC to sets of experiments in which one experimental parameter, such as the concentration of NO, is varied demonstrates the potential for mass spectral clustering to elucidate the chemical factors that drive changes in the thermal properties of OA particles. Further quantitative interpretation of the thermograms of the mass spectral clusters will allow for a more comprehensive understanding of the thermochemical properties of OA particles.
UR - http://www.scopus.com/inward/record.url?scp=85082409549&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85082409549&partnerID=8YFLogxK
U2 - 10.5194/acp-20-2489-2020
DO - 10.5194/acp-20-2489-2020
M3 - Article
AN - SCOPUS:85082409549
VL - 20
SP - 2489
EP - 2512
JO - Atmospheric Chemistry and Physics
JF - Atmospheric Chemistry and Physics
SN - 1680-7316
IS - 4
ER -