Three-dimensional factorization of size-resolved organic aerosol mass spectra from Mexico City

[摘要] A size-resolved submicron organic aerosol composition dataset from ahigh-resolution time-of-flight mass spectrometer (HR-ToF-AMS) collected inMexico City during the MILAGRO campaign in March 2006 is analyzed using3-dimensional (3-D) factorization models. A method for estimating theprecision of the size-resolved composition data for use with thefactorization models is presented here for the first time. Two 3-D models areapplied to the dataset. One model is a 3-vector decomposition (PARAFACmodel), which assumes that each chemical component has a constant sizedistribution over all time steps. The second model is a vector-matrixdecomposition (Tucker 1 model) that allows a chemical component to have asize distribution that varies in time. To our knowledge, this is the firstreport of an application of 3-D factorization models to data from fastaerosol instrumentation, and the first application of this vector-matrixmodel to any ambient aerosol dataset. A larger number of degrees of freedomin the vector-matrix model enable fitting real variations in factor sizedistributions, but also make the model susceptible to fitting noise in thedataset, giving some unphysical results. For this dataset and model, morephysically meaningful results were obtained by partially constraining the factor massspectra using a priori information and a new regularization method. We findfour factors with each model: hydrocarbon-like organic aerosol (HOA),biomass-burning organic aerosol (BBOA), oxidized organic aerosol (OOA), anda locally occurring organic aerosol (LOA). These four factors havepreviously been reported from 2-dimensional factor analysis of thehigh-resolution mass spectral dataset from this study. The sizedistributions of these four factors are consistent with previous reports forthese particle types. Both 3-D models produce useful results, but thevector-matrix model captures real variability in the size distributions thatcannot be captured by the 3-vector model. A tracer m/z-based method provides auseful approximation for the component size distributions in this study.Variation in the size distributions is demonstrated in a case study day witha large secondary aerosol formation event, in which there is evidence forthe coating of HOA-containing particles with secondary species, shifting theHOA size distribution to larger particle sizes. These 3-D factorizationscould be used to extract size-resolved aerosol composition data forcorrelation with aerosol hygroscopicity, cloud condensation nuclei (CCN),and other aerosol impacts. Furthermore, other fast and chemically complex 3-Ddatasets, including those from thermal desorption or chromatographicseparation, could be analyzed with these 3-D factorization models.Applications of these models to new datasets requires careful constructionof error estimates and appropriate choice of models that match theunderlying structure of those data. Factorization studies with these 3-Ddatasets have the potential to provide further insights into organic aerosolsources and processing.

[学科分类] 几何与拓扑