AAAR 34th Annual Conference
October 12 - October 16, 2015
Hyatt Regency
Minneapolis, Minnesota, USA
Abstract View
A Novel Strategy for Long-term Source Apportionment of Aerosol Mass Spectra
Francesco Canonaco, Kaspar Daellenbach, Imad El Haddad, Monica Crippa, JAY SLOWIK, Yuliya Sosedova, Carlo Bozzetti, Ru-Jin Huang, Urs Baltensperger, Christoph Hueglin, Hanna Herich, Andre Prévôt, Paul Scherrer Institute
Abstract Number: 299 Working Group: Source Apportionment
Abstract Positive matrix factorization (PMF) analysis of organic aerosol mass spectra provides important information regarding ambient aerosol sources. Investigation of long-term data presents special challenges, as the chemical fingerprint of primary and secondary sources can evolve in response to seasonal changes. Here we present a new method for PMF analysis of long term datasets using the multilinear engine. PMF is applied to a 4-week rolling window across the dataset, with a step size of one day. For each PMF analysis, exploration of the solution space and selection of the optimal solution are executed automatically according to predefined criteria that can be adapted to the studied dataset. This automatic rolling window is implemented within the SoFi toolkit (“AuRo-SoFi”) and applied to one year of organic aerosol mass spectra collected by an aerosol chemical speciation monitor (ACSM) Zürich from February 2011 to February 2012.
The model is constrained using the a-value approach to return three primary organic aerosol factors: hydrocarbon-like organic aerosol (HOA), cooking-influenced organic aerosol (COA), and biomass burning organic aerosol (BBOA). In addition, 1-2 oxygenated organic aerosol (OOA) factors related to secondary aerosol formation were retrieved. Each PMF analysis (i.e. window centered at a given time point) consists of 2 sets of PMF runs respectively assuming 1 or 2 OOA factors, where each set consists of 100 PMF runs in which the a-values for the primary factors a-values are varied. The optimal solution is selected automatically according to criteria such as correlation with external tracers and physically reasonable diurnal patterns. Results indicate an enhanced biomass burning fraction in winter and a reduction in OOA oxygenation in summer. We will discuss seasonal trends in factor profiles and contributions. Additionally, the AuRo-SoFi approach allows for 28 replicate PMF analyses of each time point, providing an estimate of the source variation and measurement error.