Open Collections

Stochastic modeling of space-time processes : an air pollution problem

University of British Columbia

This study presents an interdisciplinary approach to an air pollution problem that takes into account the physical bases that govern the processes of interest under the framework of a Bayesian spatiotemporal model. Based on this approach we have developed a physically motivated stochastic model that decomposes the ground-level pollutant concentration field in three components, namely: transport, local production, and largescale mean trend mostly dominated by the emission rates. The model highlighted the importance of simultaneous considering different aspects of an air pollution problem. This approach is novel in the field of environmental spatial statistics and gives a different perspective on the subject. Based on the advection equation we have modeled the transport component. The advection equation implies that two important factors have to be considered when modeling the transport of an air pollutant namely: wind field and the pollutant gradients. The equation also shows that these two factors should be considered concurrently. The local production component ignores contributions from neighboring sites and takes into account only specific local meteorological conditions. This approach reflects the fact that physics and chemistry do not depend on location and time. As a result the local production spatial structure is not modeled directly, but is inherent in the covariate fields. To specify the large-scale mean process, which is mostly dominated by the emissions rates, we used the first eigenfunction of a Principal Component Analysis performed on wind pre-filtered dataset. This new approach allowed us to capture very complex spatial patterns and to avoid using emissions inventory data. The emissions rates reveal the different capacities of a pollutant production thoughout the domain of interest. The proposed models were applied to a Lower Fraser Valley dataset and were able to quantify the transport and local contributions to ozone fields. The models also capture a number of key features of ozone behavior and show excellent agreement with the observed data and outperformed considerably simpler univariate time series models by a factor of 2 to 3.

Thesis/Dissertation

application/pdf

2009-04-22

Attribution-NonCommercial-NoDerivatives 4.0 International

http://hdl.handle.net/2429/7457

Atmospheric Science

University of British Columbia

UBCV

http://creativecommons.org/licenses/by-nc-nd/4.0/