Applications of semi-empirical and statistical techniques in urban air pollution modelling

ResearchSpace/Manakin Repository

Show simple item record

dc.contributor.advisor Singhal, N en
dc.contributor.advisor Dirks, K en
dc.contributor.advisor Salmond, J en
dc.contributor.advisor Austin, G en Elangasinghe, Madhavi en 2014-11-11T00:34:34Z en 2014 en
dc.identifier.citation 2014 en
dc.identifier.uri en
dc.description.abstract This thesis explores semi-empirical and statistical modelling techniques that are primarily based on routinely-available air quality and meteorological data for answering scientific questions related to air pollution, including the relative contribution of different sources, and the effect of meteorology on observed concentrations. The current approaches to answering these questions require either detailed emission inventories or elemental analysis of a large number of samples. Hence, they have a limited applicability in many practical situations. The results of the study suggest that the Site Optimized Semi Empirical model (SOSE) is a useful tool for identifying the most representative wind sites for air pollution applications, estimating the contribution of a major roadway from background concentrations (source apportionment), and for estimating the diurnal variation in the mixing height governing the site-specific pollutant dispersion mechanisms. The source apportionment and mixing height estimations are validated using actual measurements. The results of the study suggest that the SOSE model is capable of explaining more than 60% of the variability in hourly average CO and NO2 concentrations using meteorological predictor variables. However, it does not perform well for the prediction of particulate matter concentrations due to its complex source characteristics. Since this approach requires only routinely-available concentrations and wind data from a single monitor, it is practically applicable to many situations. As an alternative, an artificial neural network model, a nonlinear modelling approach, has been shown to produce reliable predictions for NO2 and CO (r2> 0.7) in topographically-complex Auckland, New Zealand, using only routinely available meteorological data. This study is further extended by exploring a novel approach based on artificial neural networks combined with kmeans clustering to explain the complex time series of PM10 and PM2.5 concentrations in a coastal site of New Zealand. The inclusion of cluster rankings as an input parameter to the ANN model showed a statistically significant (p<0.005) improvement to the performance of the ANN time series model and also showed better performance in detecting the occurrence of high concentrations. When modelled and observed concentrations are compared, the new approach improved the correlation coefficient (r) from 0.77 to 0.79 for PM2.5 and from 0.63 to 0.69 for PM10, and reduced the root mean squared error (RMSE) from 5.00 to 4.74 for PM2.5 and from 6.77 to 6.34 for PM10. The techniques presented here enable scientists to obtain an understanding of potential sources and their transport characteristics prior to the implementation of costly chemical analysis techniques or the implementation of advanced air dispersion models. en
dc.publisher ResearchSpace@Auckland en
dc.relation.ispartof PhD Thesis - University of Auckland en
dc.rights Items in ResearchSpace are protected by copyright, with all rights reserved, unless otherwise indicated. Previously published items are made available in accordance with the copyright policy of the publisher. en
dc.rights.uri en
dc.title Applications of semi-empirical and statistical techniques in urban air pollution modelling en
dc.type Thesis en The University of Auckland en Doctoral en PhD en
dc.rights.holder Copyright: The Author en
dc.rights.accessrights en
pubs.elements-id 460505 en
pubs.record-created-at-source-date 2014-11-11 en

Full text options

This item appears in the following Collection(s)

Show simple item record


Search ResearchSpace

Advanced Search