| Estimating Regional Spatial and Temporal Variability of PM2.5 Concentrations Using Satellite Data, Meteorology, and Land Use Information Yang Liu,1,* Christopher J. Paciorek,2 and Petros Koutrakis1 1Department of Environmental Health, and 2Department of Biostatistics, Harvard University, School of Public Health, Boston, Massachusetts, USA Abstract
Background: Studies of chronic health effects due to exposures to particulate matter with aerodynamic diameters ≤ 2.5 µm (PM2.5) are often limited by sparse measurements. Satellite aerosol remote sensing data may be used to extend PM2.5 ground networks to cover a much larger area. Objectives: In this study we examined the benefits of using aerosol optical depth (AOD) retrieved by the Geostationary Operational Environmental Satellite (GOES) in conjunction with land use and meteorologic information to estimate ground-level PM2.5 concentrations. Methods: We developed a two-stage generalized additive model (GAM) for U.S. Environmental Protection Agency PM2.5 concentrations in a domain centered in Massachusetts. The AOD model represents conditions when AOD retrieval is successful ; the non-AOD model represents conditions when AOD is missing in the domain. Results: The AOD model has a higher predicting power judged by adjusted R2 (0.79) than does the non-AOD model (0.48) . The predicted PM2.5 concentrations by the AOD model are, on average, 0.8–0.9 µg/m3 higher than the non-AOD model predictions, with a more smooth spatial distribution, higher concentrations in rural areas, and the highest concentrations in areas other than major urban centers. Although AOD is a highly significant predictor of PM2.5, meteorologic parameters are major contributors to the better performance of the AOD model. Conclusions: GOES aerosol/smoke product (GASP) AOD is able to summarize a set of weather and land use conditions that stratify PM2.5 concentrations into two different spatial patterns. Even if land use regression models do not include AOD as a predictor variable, two separate models should be fitted to account for different PM2.5 spatial patterns related to AOD availability. Key words: AOD, GAM, GASP, GOES, PM2.5, RUC, satellite aerosol remote sensing, spatial synoptic classification. Environ Health Perspect 117:886–892 (2009) . doi:10.1289/ehp.0800123 available via http://dx.doi.org/ [Online 28 January 2009]
Address correspondence to Y. Liu, Department of Environmental and Occupational Health, Rollins School of Public Health, Emory University, 1518 Clifton Rd. NE, Atlanta GA 30322 USA. Telephone: (404) 727-2131. Fax: (404) 727-8744. E-mail: yang.liu@emory.edu *Current address: Department of Environmental and Occupational Health, Rollins School of Public Health, Emory University. We thank S. Kondragunta, C. Xu, and P. Ciren at the National Oceanic and Atmospheric Administration National Environmental Satellite, Data, and Information Service ; S. Sheridan of Kent State University ; and M. Franklin and J. Yanosky of Harvard University for their technical support. The study is supported by the Harvard–U.S. Environmental Protection Agency Center on Particle Health Effects (R-827353 and R-832416) and by the Health Effects Institute (HEI) (4746-RFA05-2/06-7) , an organization jointly funded by the U.S. Environmental Protection Agency (EPA) (Assistance Award No. R-82811201) and certain motor vehicle and engine manufacturers. The contents of this article do not necessarily reflect the views of HEI, or its sponsors, nor do they necessarily reflect the views and policies of the U.S. EPA or motor vehicle and engine manufacturers. The authors declare they have no competing financial interests. Received 24 August 2008 ; accepted 28 January 2009.
The full version of this article is available for free in HTML or PDF formats. | 
