AbstractsPsychology

The first part of this thesis is about the VE3SPA project. Land use regression (LUR) models are often used to predict the outdoor air pollution at the home address of study participants, to study long-term effects of air pollution. While several studies have documented that PM2.5 mass measured at a central site correlates well in time with personal exposure, little is known about how well spatial variation of home outdoor concentration predictions represent personal exposure. For this project, outdoor and indoor concentrations were measured at 15 participants in Utrecht (the Netherlands), Helsinki (Finland) and Barcelona (Spain). Simultaneously, the personal exposure was measured for these 45 participants using pump units in small backpacks. Measurements were conducted for 6 times 96 hours (Monday- Friday) in three different seasons (winter, summer and spring/autumn). The aim was to assess the association between LUR predicted home outdoor concentrations and measured personal exposure of particulate matter with a diameter of <2.5 µm (PM2.5), soot, NO2, NOx and the elemental composition of PM2.5. The LUR models were developed by the ESCAPE project, a large European project studying the long-term effects of air pollution. The models were better able to predict the personal exposure for components with fewer indoor sources, such as soot. Soot LUR models explained 39%, 44% and 20% of personal exposure variability (R2) in Helsinki, Utrecht and Barcelona. This illustrates that predicting long-term exposure of study participants for epidemiological studies remains challenging. The second aim was to assess the temporal association of PM2.5 elemental composition measured at a central site with measured personal exposure. The generally high correlations that were found support the use of a central site for assessing exposure of PM components in time series studies for most elements. In the second part of this thesis models for ultrafine particles (UFP) and black carbon (BC) were developed. Currently, LUR models have been developed extensively for NO2 and PM2.5, but not for ultrafine particles (UFP). This project was a short-term (mobile) measurement campaign, using an electric car to transport the equipment. To develop the models measurements were done for 30 minutes at 80/81 sites in Rotterdam and Amsterdam (the Netherlands), respectively. These measurements were repeated in three seasons. LUR models for UFP and BC were developed using the GIS predictors and methodology from the ESCAPE study. The percentage explained variability (R2) varied between 0.34-0.50 for BC and 0.32-0.43 for UFP. Traffic variables were present in every model. The LUR models for UFP and especially BC predicted spatial contrasts from external datasets, derived from previous campaigns based on longer sampling durations, well. In conclusion, this short-term sampling campaign for UFP and BC delivered fairy robust models.