Source-Receptor Modeling of Air Pollution : Assessment of Source Contributions: Source Characterization and Chemometric Applications

University dissertation from Stockholm : Institutionen för tillämpad miljövetenskap (ITM)

Abstract: Particles released to the atmosphere from anthropogenic sources affect the Earth’s climate as well as the health of the population. Anthropogenic sources of atmospheric particles are e.g. the combustion of biomass and fossil fuels, road, brake and tire wear and various industrial activities. There is great interest to find the importance of different sources to the particle concentrations in the atmosphere to minimize their impact on climate and health. This work investigates how well particle sources are assessed when using source-receptor models. The main focus of this work lies in retrieving the uncertainties and difficulties of using these models on sampled particulate data. Comparison is made with other methods, e.g. a meteorological air pollution dispersion model and a tracer method.Depending on the source origin, the uncertainties of the source profiles are different and require different approaches to be quantified. Conclusions drawn from this thesis are:• assessment of point-sources which affect the sampling site less frequently, requires longer periods of sampling. • at a sampling site impacted by many different sources the time resolution of the samples has to be high in order to enable a differentiation between the sources. • natural sources, or area sources that impact the sampling site frequently, requires fewer samples and can be assessed even when the time resolution of the samples is as low as 3-4 days. This is further stressed by the fact that the natural sources are well characterized in terms of inorganic compounds.Using levoglucosan as sole quantitative tracer for domestic wood burning was shown to be associated with large uncertainties. In contrast to the unique tracer method for source assessment, the multivariate methods will also point at uncertainties in the data model, when the model cannot give a good estimate from the sampled data.

  This dissertation MIGHT be available in PDF-format. Check this page to see if it is available for download.