Articles | Volume 15, issue 14
Atmos. Chem. Phys., 15, 8077–8100, 2015
Atmos. Chem. Phys., 15, 8077–8100, 2015

Research article 22 Jul 2015

Research article | 22 Jul 2015

Mapping gas-phase organic reactivity and concomitant secondary organic aerosol formation: chemometric dimension reduction techniques for the deconvolution of complex atmospheric data sets

K. P. Wyche1,2, P. S. Monks2, K. L. Smallbone1, J. F. Hamilton3, M. R. Alfarra4,5, A. R. Rickard3,6, G. B. McFiggans4, M. E. Jenkin7, W. J. Bloss8, A. C. Ryan9, C. N. Hewitt9, and A. R. MacKenzie10 K. P. Wyche et al.
  • 1Air Environment Research, School of Environment and Technology, University of Brighton, Brighton, BN2 4GJ, United Kingdom
  • 2Department of Chemistry, University of Leicester, Leicester, LE1 7RH, United Kingdom
  • 3Wolfson Atmospheric Chemistry Laboratories, Department of Chemistry, University of York, York, YO10 5DD, United Kingdom
  • 4School of Earth, Atmospheric and Environmental Sciences, University of Manchester, M13 9PL, United Kingdom
  • 5National Centre for Atmospheric Science, University of Manchester, Manchester, M13 9PL, United Kingdom
  • 6National Centre for Atmospheric Science, University of York, York, YO10 5DD, United Kingdom
  • 7Atmospheric Chemistry Services, Okehampton, Devon, EX20 1FB, United Kingdom
  • 8School of Geography, Earth and Environmental Sciences, University of Birmingham, Birmingham, B15 2TT, United Kingdom
  • 9Lancaster Environment Centre, Lancaster University, Lancaster, LA1 4YQ, United Kingdom
  • 10Birmingham Institute of Forest Research, University of Birmingham, Birmingham, B15 2TT, United Kingdom

Abstract. Highly non-linear dynamical systems, such as those found in atmospheric chemistry, necessitate hierarchical approaches to both experiment and modelling in order to ultimately identify and achieve fundamental process-understanding in the full open system. Atmospheric simulation chambers comprise an intermediate in complexity, between a classical laboratory experiment and the full, ambient system. As such, they can generate large volumes of difficult-to-interpret data. Here we describe and implement a chemometric dimension reduction methodology for the deconvolution and interpretation of complex gas- and particle-phase composition spectra. The methodology comprises principal component analysis (PCA), hierarchical cluster analysis (HCA) and positive least-squares discriminant analysis (PLS-DA). These methods are, for the first time, applied to simultaneous gas- and particle-phase composition data obtained from a comprehensive series of environmental simulation chamber experiments focused on biogenic volatile organic compound (BVOC) photooxidation and associated secondary organic aerosol (SOA) formation. We primarily investigated the biogenic SOA precursors isoprene, α-pinene, limonene, myrcene, linalool and β-caryophyllene. The chemometric analysis is used to classify the oxidation systems and resultant SOA according to the controlling chemistry and the products formed. Results show that "model" biogenic oxidative systems can be successfully separated and classified according to their oxidation products. Furthermore, a holistic view of results obtained across both the gas- and particle-phases shows the different SOA formation chemistry, initiating in the gas-phase, proceeding to govern the differences between the various BVOC SOA compositions. The results obtained are used to describe the particle composition in the context of the oxidised gas-phase matrix. An extension of the technique, which incorporates into the statistical models data from anthropogenic (i.e. toluene) oxidation and "more realistic" plant mesocosm systems, demonstrates that such an ensemble of chemometric mapping has the potential to be used for the classification of more complex spectra of unknown origin. More specifically, the addition of mesocosm data from fig and birch tree experiments shows that isoprene and monoterpene emitting sources, respectively, can be mapped onto the statistical model structure and their positional vectors can provide insight into their biological sources and controlling oxidative chemistry. The potential to extend the methodology to the analysis of ambient air is discussed using results obtained from a zero-dimensional box model incorporating mechanistic data obtained from the Master Chemical Mechanism (MCMv3.2). Such an extension to analysing ambient air would prove a powerful asset in assisting with the identification of SOA sources and the elucidation of the underlying chemical mechanisms involved.

Short summary
This paper describes a new ensemble methodology for the statistical analysis of atmospheric gas- & particle-phase composition data sets. The methodology reduces the huge amount of data derived from many chamber experiments to show that organic reactivity & resultant particle formation can be mapped into unique clusters in statistical space. The model generated is used to map more realistic plant mesocosm oxidation data, the projection of which gives insight into reactive pathways & precursors.
Final-revised paper