Abstract
Exposure assessment is one of the key issues for health effect estimates in environmental epidemiology. Recent interest has increased in exposure modeling incorporating Geographic Information System (GIS) data to capture small-scale spatial variability in air pollution concentrations. Land use regression (LUR) modeling is one of the most popular models due
... read more
to the high resolution mapping technique. However, the performances of the LUR models have not been well explored. Although several studies have shown evidence of acute effects of particulate matter on cardiovascular mortality, for components results were inconsistent. Very few studies have assessed health effects related to long-term exposure to elemental composition in particulate matter. Lack of spatially resolved elemental composition measurement data and a lack of models for elemental composition have contributed to this gap. The research is conducted within the framework of the European Study of Cohorts for Air Pollution Effects (ESCAPE) study which aims at qualifying chronic health effects of outdoor air pollution in multiple cities of Europe. The aim of this thesis is: 1. To evaluate the performances of LUR models in terms of model fit and prediction ability 2. To develop LUR models for particle compositions 3. To estimate associations between long-term exposure to particle compositions and cardiovascular mortality Evaluation is an essential part of LUR model development. Our results suggested that: 1) the prediction ability of a LUR model to independent places is overestimated when the number of sampling sites is small. Therefore, truly independent evaluation data are especially useful when LUR models are developed from small training sets. This conclusion is applicable in many European cities of the ESCAPE study areas though the LUR models still explained a substantial fraction of variation of concentrations measured at independent sites. 2) It is possible to develop LUR models for nitrogen dioxide and particulate matter using combined study areas which are able to predict independent sites and new areas reasonably well. 3) Elemental compositions of particulate matter in the ESCAPE study areas can be successfully developed using the LUR modeling strategy which results in good model fit for non-tailpipe emissions elements such as Copper, Iron and Zinc and poor performance for Silicon, Vanadium and Potassium. Using the estimates of LUR models, we found no evidence of long-term effects of the exposures of ambient elemental compositions of particulate matters (Copper, Iron, Potassium, Nickel, Sulfur, Silicon, Vanadium and Zinc) on cardiovascular mortality in 19 European cohorts of the ESCAPE study
show less
