Collins, John Philip (2008) Deconstructing adsorption variability: the prediction of spatial uncertainty in pollutant movement. Doctoral thesis, Durham University.
Land pressures today and government policy-requires previously developed, 'brownfield' land to be brought back into beneficial use. The nature of these sites means that they may have been subject to some form of contamination from previous uses. The risk any pollutant has to human health and the environment must be assessed and, if deemed unacceptable, remediation must be undertaken. Risk assessment may be carried out utilising generic values for contaminant properties that can give misleading results. This thesis describes the effort to further assess the controls on adsorption of organic pollutants and its spatial variability. Spatial sampling of two brownfield sites was undertaken with generic soil parameters being measured. To better describe soil organic matter, organic extracts were prepared from soils, allowing NMR spectra to be collected. The collected soil dataset is analysed to discern any correlations between soil parameters. The nature of the organic pollutants used in this study (benzene, phenol, p-xylene and p-cresol) is described using calculated molecular descriptors. The variation in experimental adsorption results, provided by Sheffield University, were then statistically analysed using soil measures as predictors and then also adding molecular descriptors to the analysis. The percentage of black carbon may also have an influence on adsorption and so this was also measured and added to the list of predictors available for inclusion in stepwise regression. Results show that adsorption of these organic compounds can be partially described using the measured soil parameters. Molecular descriptors such as a molecule's surface area can also be used to predict adsorption. The percentage black carbon was an important predictor in only one instance for p-xylene adsorption. Soil parameters were also shown to be predicted by other soil variables from the dataset, giving good results that were improved upon by transforming all parameters to normality.
|Item Type:||Thesis (Doctoral)|
|Award:||Doctor of Philosophy|
|Copyright:||Copyright of this thesis is held by the author|
|Deposited On:||08 Sep 2011 18:28|