The analysis of aerosol time-of-flight mass spectrometry data

Aerosol particles have received significant public and scientific attention in recent years due to studies linking them to global climatic changes and human health effects. In 1994, Prather et al. developed Aerosol Time-of-Flight Mass Spectrometry (ATOFMS), the first technique capable of rapidly determining the precise sizes and chemical compositions of many individual heterogeneously sized aerosol particles in real-time. With the technique developed to analyze aerosol particles, the new challenge has become the analysis of ATOFMS data.; ATOFMS data take the form of a mass spectrum and aerodynamic diameter of each particle. There are fundamentally two approaches to extracting useful information from these data; the particles can be associated by like kind based on their mass spectra or their mass spectra can be used to determine the chemical comp6sition of the particle population. Successful approaches to these two tasks are presented.; The artificial neural network algorithm, ART-2a, is used to classify particles by their chemical characteristics. This allows researchers to cluster all similar particles with no prior knowledge of the particles present at the time. This method and the careful analysis of the particles contained within the clusters are used to observe changes in the proportions of ambient aerosol particles from various sources in Long Beach, California. The results of an ART-2a analysis are scaled quantitatively using methods and coefficients developed by Allen et al. and are used to compare the particle populations at two different field sites.; Two very different approaches are taken towards the quantification of chemical species present within a particle population. A univariate approach is used to determine the mass concentrations of ammonium and nitrate ions in the ambient atmosphere with hourly time resolution. A multivariate approach is used to determine the mass concentrations of 44 atmospheric chemicals, but only during isolated four-hour time blocks. In the latter case, no human intervention is required in the analysis. It is hoped that the studies presented here will form the foundation for a new generation of approaches to the analysis of ATOFMS data.