Spatial Analysis of Air Pollution and Development of a Land-use Regression (LUR) Model in an Urban Airshed

Abstract

P-659 Introduction: The Detroit Children's Health Study is an epidemiologic study examining associations between chronic ambient environmental exposures to gaseous air pollutants and respiratory health outcomes among elementary school-age children in an urban airshed. The exposure component of this study was to use ambient and GIS data from selected school sites to assess spatial variability and develop a land-use regression (LUR) model that could estimate intra-urban gradients of ambient air pollution levels for use in the health assessment. Methods: Rather than choosing an arbitrary regression model, a statistical-based approach was used to develop the model in which the GIS variables and school sites were selected to span the mathematical space for all schools and variables considered within the urban study area. Passive exposure monitoring methods for volatile organic compounds (VOC) and nitrogen dioxide were used at the selected schools for weeklong integrals to mimic chronic exposures. Monitoring was performed for 6 weeks during the summer of 2005 when there was greater occurrence of stable air masses and low winds to assess spatial gradients. Results: Based on a correlation analysis of the 45 variables originally considered by type, the following variables were chosen as predictor variables in the model: distance to the nearest road with traffic volume of at least 50,000 vehicles per day, distance to the nearest road with traffic volume of at least 90,000 vehicles per day, traffic intensity within 1000 m, population density within 500 m, distance to the nearest large total VOC emitting facility, distance to the nearest large PM2.5 emitting facility, distance to the nearest large manganese emitting facility, and distance to the nearest border crossing. Kruskal-Wallis tests were performed on common VOCs such benzene, toluene, ethylbenzene, and xylene isomers (collectively known as BTEX) and 1,3-butadiene by enumeration district (second digit of census tract number). Tests revealed overall differences at 10% significance level among enumeration districts for benzene. Discussion and Conclusions: Preliminary results indicate coarse spatial gradients exist for select VOCs within the urban study area. This suggests variability throughout the area that would be conducive to development of a spatial model. Development of the LUR model using the predictor variables will be presented. Disclaimer- Although this work was reviewed by EPA and approved for publication, it may not necessarily reflect official Agency policy.