Background
PartI
PartII
PartIII
Summary
Acknowledgement

Background

        Land use regression (LUR) models have been increasingly used as a cost-effective approach to estimate outdoor air pollution concentrations. This method uses measurements of pollutants at multiple sites, and potentially associated geographic attributes (e.g. land use, population density, and traffic patterns) in a Geographic Information System (GIS), to build regression models which can be used to predict air pollution concentrations at unmeasured areas. Stochastic modeling is used to determine which predictor variables best explain the pollution concentrations measured at a number of network locations.1

        In Metro Vancouver, a LUR mode for NO2 was developed in 20032. As part of my thesis, it has been recently updated to 2009/10. The methodology for developing the 2009/10 largely followed those adopted in developing the 2003 model, with up-to-date input variables. Briefly, ambient concentrations of NO2 were sampled at 116 sites across Metro Vancouver. For each of the 116 measurement sites, potentially associated variables were generated in GIS and linear regression models were built with the most predictive covariates. The regression equations were then rendered as maps using algebraic features under spatial analyst in GIS2. Maps rendered from the two models are shown below, for 2003 (left) and 2009/10 (right) respectively.
Modeled NO2 surface in 2003

Modeled NO2 surface for 2009/10

        Currently, the main application of LUR model is to estimate subjects’ exposure in large epidemiological studies.  Once a LUR model is developed, exposure can be estimated by geocoding addresses of interests, i.e. subjects' homes, schools etc, into latitude/longitude coordinates and determining the modeled pollutant concentrations at those locations.2  However, apart from this application, other potential uses of the LUR model haven’t been fully explored.  This project was intended to explore other applications of the LUR model, air quality evaluation in particular. In addition, spatial autocorrelation will be examined as part of the model validation.

Objectives

Specifically, the objectives of this project include: 

  1. To assess the change in spatial patterns of NO2 from 2003 to 2009/10 
  2. To identify areas where the ambient NO2 concentration exceeds Metro Vancouver’s annual objective (21ppb)3 
  3. To test the assumption that sampling sites are independent using spatial autocorrelation (Moran’s I)

Because of the different aspects addressed in the three objectives, they will be presented in three parts individually, each with introduction, methods, results and discussion/conclusion of its own.

___________________________

1.  Hoek G, Beelen R, de Hoogh K, Vienneau D, Gulliver J, Fischer P, Briggs D: A review of land-use regression models to assess spatial variation of outdoor air pollution. Atmospheric Environment 2008, 42(33):7561-7578.

2.  Henderson SB, Beckerman B, Jerrett M, Brauer M: Application of land use regression to estimate long-term concentrations of traffic-related nitrogen oxides and fine particulate matter.Environmental Science & Technology 2007, 41(7):2422-2428. 
3.  Metro Vancouver. 2009 Lower Fraser Valley Air Quality Report. June, 2010

Content last updated: Dec 9, 2010